Introduction

Subduction zones are powerful physical and chemical engines where sediments, crust, and lithosphere sink and, on the way down, interact with asthenospheric mantle. This equilibrates Earth’s mantle with its volatile- and incompatible-element-rich outer layers and produces magmas. Understanding this deep recycling system is a scientific challenge requiring multiple approaches. One valuable perspective comes from understanding how and why lava compositions vary with depth to the subduction zone, seismically defined as the Wadati-Benioff Zone (WBZ). In particular, studying lava compositions in cross-chains, oriented perpendicular to the strike of the arc magmatic front and parallel to the WBZ dip direction, should reveal how contributions from the subducted slab and sediments and overriding mantle contributions change with increasing WBZ depth.

From the earliest days of plate tectonics, it was known that magmas generated over deeper parts of a subduction zone contained more of a typical incompatible element (potassium) than melts generated over shallower parts of the same subduction zone (Hatherton and Dickinson 1969). Hatherton and Dickinson (1969) were undecided as to the cause of this systematic variation, suggesting that it could reflect either differing stabilities of K-bearing phases in the subducted slab (i.e., amphibole vs. phlogopite) or different ascent paths or melting processes. The ‘K--h’ relationship---which we now should expand to include more of the periodic table, including large ion lithophile (LIL) elements Rb and Ba and the light Rare Earth elements (LREE)---is still a robust constraint, but it remains unclear what this fundamental relationship reveals about subduction zone processes. We now know that nearly all of the major elements in most arc magmas come from partial melting of convecting asthenosphere above the subducting slab, and that LIL and other incompatible element abundances are due to distillation of subducted sediments (Plank and Langmuir 1993) or dehydrating oceanic crust (Ishikawa and Tera 1999). Two attractive endmember possibilities are that the K--h relationship indicates a greater role for sediment melting with greater WBZ depth, or that lower degrees of mantle melting occur as distance increases from the magmatic front. These simplified possibilities provide a useful frame for the themes addressed in this paper.

The significance of the K--h relationship can be most confidently explored using intra-oceanic arc (IOA) systems, those convergent margins built on oceanic lithosphere. Like most convergent margin magmas, IOA magmas are generally fractionated and so are affected by assimilation and fractionation. However, IOA assimilation involves mafic arc crust that is chemically and isotopically similar to the arc magmas that pass through it; furthermore this crust is more refractory than continental crust. Ergo, assimilation and its effects on magma compositions are minimized in IOA systems, particularly when compared to melts that transit the thicker and more reactive crust of continental magmatic arcs. Furthermore, involvement of lithosphere in IOA magmagenesis is unlikely to obscure melt compositions, because recently formed IOA lithosphere should not have the ancient isotopic heterogeneities and LIL enrichments of ancient subcontinental lithosphere. Finally, IOA cross-chains built on back-arc basin crust should show an absolute minimum of contamination because crust beneath such cross-chains changes from relatively thick (~20--30 km) along the magmatic front to quite thin behind the arc (~5--10 km).

Here we explore how subducted lithosphere and sediments interact with the overlying mantle wedge and make it melt. New petrologic, geochemical, and isotopic results are reported for a previously unstudied cross-chain in the southern part of the 3,500 km long--Izu--Bonin--Mariana (IBM) arc system (Fig. 1a, b). We use these results to examine how processes of melt generation vary with increasing depth to the subduction zone. To the best of our knowledge, this is the first such report for a cross-chain in an IOA setting associated with an actively spreading back-arc basin. (The Kasuga cross-chain to the north (Fig. 1b; Fryer et al. 1997; Stern et al. 1993) is associated with a portion of the Mariana Trough that is rifting, not spreading (Martinez et al. 1995).) Our most important conclusion is that sediment fluids or melts are not responsible for cross-chain chemical and isotopic variations but that supercritical fluids released from serpentinites and interacting with modestly enriched mantle might be.

Fig. 1
figure 1

a Locality map of the Izu (I)–Bonin (B)–Mariana (M) arc system. b Tectonic setting of the Mariana arc system (modified after Martinez and Taylor 2003). The Mariana Trough backarc basin spreading axis and rifts are shown as heavy lines. The Pacific Plate subducts beneath the Mariana arc at the Mariana Trench (dotted line), where the 6,000 m isobath is shown. The 100, 200 and 500 km contours of the Wadati-Benioff Zone (WBZ) are shown as labeled dashed lines. The magmatic arc (including the island of Guguan) lies about 130 km above the WBZ. Water depths shallower than 3,000 m are shown in grey. Black box shows the location of (d). K shows the location of the Kasuga cross-chain. c Cross-sectional view of the central Mariana subduction system using the earthquake catalog of Engdahl et al. (1998). Black circles denote earthquake hypocenters in a 50 km wide volume centered at 17.33°N latitude. d HMR-1 sonar backscatter image of the Guguan cross-chain, showing the location of dredges sampling carried out during Cook 7 expedition. Note location of Guguan island

Full size image

Geologic setting

The Mariana magmatic front in this study is represented by Guguan island and a small submarine cone to the north of Guguan (N. Guguan) sampled during the 2001 Cook 7 expedition by dredge 50 (17°24.2′N, 145°51.8′E, 1,928–1,695 m). The Guguan cross-chain extends due west from Guguan and trend perpendicular to the magmatic front. Guguan and N. Guguan lie about 125 km above the WBZ (Fig. 1c), similar to the 115±10 km mean depth for Mariana arc volcanoes (England et al. 2004). There are two rear-arc volcanoes in the cross-chain that we were able to sample during Cook 7: West Guguan (W. Guguan) and Guguan 2. The W. Guguan edifice lies on the western submerged flank of Guguan, about 150 km above the WBZ, and was sampled by Cook 7 D49 (17°15.2′N, 145°42.1′E, 1,890--1,708 m). The Guguan 2 edifice lies over a very steep part of the WBZ, with an estimated depth of 230 km (Fig. 1c). These depth estimates are approximate, but it is clear that Guguan and N. Guguan lie above shallower sections of the WBZ than do W. Guguan and Guguan 2. Crust outboard of the trench is Jurassic (Stern et al. 2003), the oldest subducted beneath any convergent margin today, so Mariana is the ‘cold subduction zone’ endmember for global subduction zones.

Cook 7 D46, 47, and 48 sampled W. Guguan. D46 (17°12.3′N, 145°35′E, 1,821–1,645 m) and D48 (17°16.5′N, 145°37.4′E, 2,248–2,160 m) recovered abundant fresh basalt. D47 (17°15.4′N, 145°33.6′E, 1,468–1,302 m) recovered a small amount of altered volcanics which were not studied. We tried but failed to sample another edifice ~25 km WNW of Guguan 2, probably due to thick Mn armoring. Locations of dredge materials discussed in this report are shown on Fig. 1d. The slowly spreading axis of the Mariana Trough back-arc basin lies about 75 km west of Guguan 2 (Fig. 1b), so Guguan 2 lies about one-third of the distance from the magmatic front to the back-arc basin spreading axis. At the latitude of the Guguan cross-chain (~17°20′N), the Mariana Trough spreads at a rate of 25–30 mm/year (Kato et al. 2003).

The history of volcanism in the Guguan cross-chain is poorly known. Guguan erupted in 1882–1884 but there are no records of any submarine eruptions. The small cone sampled by D50 appears to be young, as evidenced by recovery of abundant fresh basalt with little Mn coating. W. Guguan may be extinct, as suggested by Mn-encrusted basalt recovered in D49. Guguan 2 appears to be relatively young, as suggested from good recovery of fresh basalts with little Mn coating, and by the fact that the sonar images show that debris aprons extending from the edifice onto the seafloor are not covered by hemipelagic sediments (Fig. 1d). There are no radiometric ages, but Guguan cross-chain volcanoes must have started to grow after seafloor spreading in the Mariana Trough began sometime prior to the start of the Gauss magnetic chron (2.6–3.58 Ma; Ishihara et al. 2001).

The Guguan cross-chain is a particularly interesting place to examine how elemental fluxes vary with increasing depth to the WBZ, because Guguan lavas are broadly accepted to be the magmagenetic endmember where the subduction component is delivered as hydrous fluid to the site of melt generation, with minimal contribution from subducted sediment (Elliott et al. 1997; Woodhead et al. 2001). Glass inclusions in Guguan olivines contain 2–4 wt% H2O and average around 3 wt% H2O (Kelley 2002). Guguan lavas show the largest 238U excesses of Mariana arc lavas (Elliott et al. 1997), which is also consistent with hydrous fluxing. Guguan lavas have high δ11B and high B/Nb, leading Ishikawa and Tera (1999) to conclude that most of the hydrous fluid was derived from altered oceanic crust. All of these observations are consistent with the inference that Guguan lavas represent the fluid-fluxed endmember of Mariana arc magmas (Elliott et al. 1997).

Analytical techniques

Minerals were analyzed with the Cameca SX-50 electron microprobe at Oregon State University using a 15 kV, 30 nA, 3 μm diameter beam for feldspar, and 15 kV, 50 nA, 1 μm diameter beam for olivine, pyroxene, and spinel. For feldspar and spinel analyses, count times were 10 s for all elements. Olivine analyses had count times of 20 s for Fe and Ni, and 10 s for Mg, Al, Si, K, Ti, Na, Ca, Cr, and Mn. Pyroxene analyses used count times of 20 s for Fe, Cr, and Si and 10 s for all other elements. To minimize the effects of Na migration, Na was analyzed first.

Major element concentrations were determined by Activation Laboratories using whole-rock powders. Trace element concentrations were determined at UTD by inductively compled plasma mass spectrometer (ICP-MS). Powders were prepared in an alumina ball mill. Samples (~200 mg) were weighed into teflon beakers and digested using HF, HClO4, HCl, and HNO3. Solutions were diluted 250 times in ~10% HNO3 and further diluted 10 times in 4% HNO3 prior to analysis. A typical analytical run of 25 samples consisted of five to six certified reference materials (CRMs), and one to two procedural blanks with the remainder being sample unknowns. CRMs included USGS standards BIR-1, W-2, BHVO-2, BCR-2, AGV-2, RGM-1, GSP-2, and G-2. Acids were distilled in-house or were purchased (Baseline grade) from Seastar Chemicals. All dilutions were prepared using 18.2 MW water from a Millipore system including an Elix reverse osmosis and electrodeionization unit and a Milli-Q Element final purification unit. Trace and rare earth elements were analyzed on a Perkin Elmer-Sciex Elan 6100 dynamic reaction cell ICP-MS (DR-ICP-MS). External calibration standards were prepared from SCP multi-element ICP standards. Samples were introduced to the ICP via an autosampler at a flow rate of 1 mL min−1. Samples were nebulized using a cross-flow nebulizer and Ryton Scott spray chamber. Typical operating conditions are 0.90 L min−1 nebulizer gas flow, 1.30 L min−1 auxiliary gas flow, 13 L min−1 plasma gas flow at an operating RF voltage of 1,100 W. The ion lens is optimized at the start of each run and the voltage is dynamically modified by the instrument software to optimize analyte signal for masses of In. To monitor drift, a calibration standard and a CRM are run after every five unknowns. Procedural blanks are typically near or below detection for the operating conditions.

87Sr/86Sr was determined using the Finnigan MAT 261 solid-source mass spectrometer at UTD. Reproducibility of 87Sr/86Sr is ±0.00004. During this study, the UTD lab obtained a mean 87Sr/86Sr=0.70803±3 for several analyses of the E&A SrCO3 standard; data reported here have been adjusted to correspond to a value of 0.70800 for the E&A standard. 143Nd/144Nd was also determined using the UTD Finnigan-MAT 261 in the dynamic multicollector mode. Calculations of εNd for samples are based on values of εNd for the UCSD standard ((15.2) and BCR ((0.16; Pier et al. 1989). A total range of ±0.00002 observed for 143Nd/144Nd of the standard (mean value = 0.511868) is taken as the analytical uncertainty for the samples. Pb was separated using the technique of Manton (1988) and isotope ratios were also determined at UTD using the MAT 261 in the static multicollector mode and corrected for fractionation using results for the NBS-981 standard analyzed under the same conditions. Total processing blanks for Sr, Nd, and Pb are <0.1, <0.3, and <0.3 ng, respectively. Hf isotopes were analyzed by MC-ICP-MS using the ThermoFinnigan Neptune at Washington State University. Multiple analyses of the JMC 475 standard over the past year yield 176Hf/177Hf=0.282146±0.000010 (2 standard deviation (SD), n=63). This number is without any bias factors applied other than the standard exponential mass bias correction. Standards within single days reproduce within a total range of about 0.000010, which is taken to approximate the analytical uncertainty. Final Hf isotopic values for samples are reported relative to 176Hf/177Hf=0.282160 for the JMC 475 Hf standard. Samples analyzed for Hf isotopes were analyzed for Lu and Hf concentrations and 176Lu/177Hf by isotope dilution using a mixed Lu--Hf tracer. The error on the 176Lu/177Hf is estimated to be 0.2%. Full details of the Lu--Hf method used at WSU, including chemistry and mass spectrometry are described in Vervoort et al. (2004).

All data for N. Guguan, W. Guguan, and Guguan 2 are new. All data for Guguan, except for EMP results, are from the literature (Elliott et al. 1997; Ishikawa and Tera 1999; Woodhead 1989; Woodhead et al. 2001).

Results

Petrography and mineral chemistry

Lavas from Guguan island have been subjected to extensive geochemical and isotopic studies but petrographically and mineralogically are relatively unstudied. These are mostly vesicular and porphyritic basalts and basaltic andesites, with abundant plagioclase phenocrysts and less abundant olivine and clinopyroxene (CPX) phenocrysts. Samples from N. Guguan (D50) have modal compositions that are similar to Guguan lavas, with 15–40% vesicles and 20–35% phenocrysts of plagioclase, CPX, and olivine (in order of decreasing abundance). Samples from W. Guguan (D49) contain 15–20% vesicles and 30–35% phenocrysts of plagioclase, clinopyroxene, and olivine. Samples from Guguan 2 are typically vesicular (1–50% vesicles) in spite of being collected from over 1,600 m deep. Guguan 2 basalts contain 1–20% phenocrysts of olivine, CPX, and plagioclase and so are less porphyritic than lavas from edifices that are closer to the magmatic front. Abundances of phenocrysts, groundmass, and vesicles are summarized in Table 1.

Table 1 Petrographic and major element data
Full size table

Phenocrysts were analyzed for representative samples for each edifice (N. Guguan: D50–1–1 and 2–1; W. Guguan: D49–1–2b; Guguan 2: D46–1–1 and D48–1-1). We also analyzed phenocrysts in two Guguan basalts (G-1 and G-2 collected by RJS in 1977). Complete analytical results are available as supplementary materials through this journal (Supplementary Tables s1, s2, s3, s4). Guguan olivine (OL) phenocrysts are Fe rich, Fo62 to Fo76, with rims as Fe rich as Fo49. These are in equilibrium with CPX, the cores of which have Mg# 64–76, also with more Fe-rich rims. OL and CPX coexist with plagioclase (PL) phenocrysts. PL phenocrysts have heterogeneous cores (An70 to An96). Phenocryst rims and groundmass feldspars are also heterogeneous, ranging from An56 to An73 (matrix) and An34 to An95 (rims). Spinels (SP) are Fe- and Ti-rich magnetites, with very low Cr contents (Cr#=100Cr/Cr+Al~0). Compared to Guguan lavas, OL from N. Guguan (D50) lavas are more magnesian (Fo74–82) and are associated with similarly magnesian CPX (Mg#=74–86), calcic PL (cores of An74–97), and titaniferous magnetite. Compared to magmatic front lavas, W. Guguan lavas contain more magnesian OL (Fo79–87) and CPX (Mg#=78–89) and calcic PL (An70–93); small amounts of OPX (Mg#=80) are also found in glomerophryric clusters.

Compared to phenocryst assemblages from along the magmatic front and from W. Guguan, Guguan 2 lavas contain more homogeneous and magnesian OL (D46–1-1: Fo 86–91; D48-1-1: Fo 80-89). Rare pyroxene phenocrysts in D48-1-1 are relatively fractionated (CPX Mg#=79–82; OPX Mg#=75–89). Groundmass PL in D46-1-1 is relatively sodic (An76–81), as are PL phenocryst cores in D48-1-1 (An77–90). Spinels in Guguan 2 lavas are also compositionally distinct from magmatic front spinels. Spinels from D46-1-1 have TiO2<0.65% and are rich in Cr, with Cr#=0.55–0.63 whereas spinels from D48-1-1 are chromites, with TiO2<0.75% and Cr# up to 0.75.

Compositional relationships between mafic phenocrysts (represented by OL) and plagioclase are shown on Fig. 2a. Convergent margin OL--PL relations are generally distinct from those of MORB and OIB. Phenocrysts in magmatic front lavas (Guguan and N. Guguan) fall in the field defined by arc lavas, whereas those from W. Guguan and Guguan 2 plot in fields for MORB and OIB. These systematic differences may be due to the rotation of the plagioclase solvus as a result of high magmatic water fugacities of arc melts, so that more calcic plagioclase is in equilibrium with Fe-rich OL than is seen for anhydrous MORB and OIB (Wagner et al. 1995). This interpretation is consistent with the high (2–4%) water contents found in Guguan glass inclusions. The similarity of W. Guguan and Guguan 2 OL-PL relations to those observed in MORB and OIB suggests that the rear-arc magmas evolved under conditions of low magmatic water fugacity, but this conclusion is difficult to reconcile with estimates of water in glass inclusions (GI) in olivine phenocrysts, as discussed later.

Fig. 2
figure 2

a Plagioclase An content vs. olivine Fo for Guguan cross-chain. Symbols mark mean mineral pair compositions for each location and error bars define 2-σ variation. Note mineral pair compositions become increasing arc-like towards the magmatic front, indicating the increasing influence of water on plagioclase composition. Arc field compiled from BSVP (1981a), Conrey et al. (1997), Dixon and Batiza (1979), Meijer and Reagan (1981) and unpublished Mariana arc data. MORB field compiled from Allan et al. (1987), BSVP (1981c), Davis and Clague (1987), Niu and Batiza (1994), O’Donnell and Presnall (1980), and Perfit and Fornari (1983). OIB field compiled from Breddam (2002), BSVP (1981b, c), and Garcia et al. (2000). b Trivalent ion plot for spinels from Guguan cross-chain. Arc tholeiite fields (enclosed by dashed line) and MORB field are from Barnes and Roeder (2001). Ti-rich magnetites from D50 and Guguan lie in the arc magnetite field and D46 and D48 Cr-spinels plot in MORB field

Full size image

Similar variations in spinel compositions are also found for the Guguan cross-chain (Fig. 2b). Guguan and N. Guguan spinels are magnetites, characteristic of arc lavas, whereas Guguan 2 spinels are similar to those of MORB (and OIB, not shown).

Temperature and pressure estimates

Magmatic temperatures were estimated using the CPX-liquid geothermobarometer of Putirka et al. (1996, 2003), the olivine-liquid geothermometer of Ford et al. (1983), and the olivine-spinel geothermometers of Fabriés (1979), Roeder et al. (1979) and Balhaus et al. (1991). Olivine--spinel pairs of olivine phenocrysts with spinel inclusions were used for D46 and D48; for Guguan lavas we used the rim compositions of spinels and olivines that crystallized separately. Estimated temperatures differ between methods; the lowest are from the olivine--spinel thermometers and the highest are from the CPX thermometer (Table 2). Only temperature estimates for sample D46 from the olivine-liquid and CPX-liquid thermometers are similar at ~1,322–1,344°C.

Table 2 Estimated magmatic temperatures and pressures
Full size table

We regard the temperatures from the mineral-liquid thermometers of Ford et al. (1983) and Putirka et al. (1996, 2003) to be more reliable than the olivine--spinel temperature estimates. One issue with the latter method is that D48 olivines are oscillatory zoned. This complicates matching equilibrium olivine and spinel compositions, as the spinel inclusions we analyzed for geothermometry are located along zone boundaries. The olivine--spinel geothermometers are also pressure sensitive. Even though the two mineral-liquid thermometers mostly do not agree (except for sample D46), both show an increase in temperature away from the magmatic front and are highest for sample D46 (Table 2). However, only in sample D46 are the olivines approximately in equilibrium with bulk rock compositions, having an olivine-liquid Kd (Fe/Mg) of 0.29–0.3. In contrast, the CPX-liquid Kd (Fe/Mg) for N. Guguan at 0.27--0.29, and the Guguan lavas at 0.24–0.26 are close to the equilibrium values of 0.27 (Putrika 1999). In addition, sample D46 approaches equilibrium with a CPX-liquid Kd (Fe/Mg) of 0.32. Based on this characteristic, the CPX-liquid geothermometer appears to be the more reliable of the two mineral-liquid thermometers for our samples. The CPX-liquid thermometers also yield reasonable magmatic temperatures (1,228–1,344°C).

Pressures were calculated with the Putrika et al. (1996, 2003) method. Estimated pressures are from 0.8 to 3.6 kbar for D50 and D49 lavas, with median pressures of 2.2 and 1.8 kbar, respectively (Table 2). In contrast, the D48 pressures were 6.3 kbar and the D46 pressures are 8.7 kbar, revealing an increasing depth of equilibration away from the magmatic front. Unfortunately, we did not have an independent method for checking our pressure estimates, such as the Ol/CPX Ca exchange geobarometer of Köhler and Brey (1990).

Major element variations

Samples analyzed from the Guguan cross-chain are mostly basalt, with subordinate basaltic andesite and andesite (Table 1; Fig. 3a). None are Ne-normative, but normative Ol is abundant in D46 samples (12–18%) and D50 samples (2–5%). D49 and most D48 samples are Qz normative. A modest K--h relationship is observed: lavas from the magmatic front and W. Guguan straddle the low-K and medium-K fields on a potassium-silica diagram, whereas Guguan 2 samples plot in the medium-K field (Fig. 3a). Guguan and N. Guguan, lying ~ 125 km above the WBZ, together have K55 (K2O contents at 55% SiO2 by linear regression) ~0.65%, whereas Guguan 2 (~230 km above the WBZ) has K55~0.9% (there is not enough spread in SiO2 for K55 to be calculated in W. Guguan lavas). This agrees with an expected increase of K55 with increasing depth to the WBZ, but the increase is less than that predicted from the global relationship (volcanoes 125 and 230 km above the WBZ are expected to have K55 ~0.9 and ~2.2%, respectively Hatherton and Dickinson 1969). The values for Guguan and N. Guguan are similar to K57.5 values of ~0.45–0.9% reported for volcanoes along the magmatic fronts of the Izu and Tonga-Kermadec arcs (Dickinson 1975). The modest increase in K with increasing WBZ depth observed for the Guguan cross-chain is similar to modest increases observed for Izu cross-chains to the north (Hochstaedter et al. 2000; Machida and Ishii 2003; Tatsumi et al. 1992). It is noteworthy that the Guguan cross-chain did not erupt shoshonites as are observed in the Kasuga cross-chain in the northern Marianas (Fryer et al. 1997).

Fig. 3
figure 3

a Potassium-silica diagram. Data for North Guguan (N. Guguan), West Guguan (W. Guguan) and Guguan 2 are from this report; data for Guguan are from Elliott et al. (1997), Ishikawa and Tera (1999), and Woodhead (1989). b Mg# vs. depth to WBZ. Note that lavas are more primitive (higher Mg#) with distance from the magmatic arc. Andesitic samples are noted. c Water contents in glass inclusions in olivines, showing mean and range for analyses (A. Shaw, written communication, 2004). There is no significant variation in water contents of cross-chain lavas

Full size image

The extent of fractionation in Guguan cross-chain lavas varies inversely with depth to the WBZ. Magmatic front lavas are strongly fractionated (Guguan basalt Mg#=40–51; N. Guguan Mg#~53), whereas those from behind the magmatic front are primitive (melts in equilibrium with mantle peridotite) to slightly fractionated (W. Guguan Mg#=61; Guguan 2 basalt Mg#=63–75). This is also observed from Ni and Cr variations, which increase from <55 ppm Ni and <100 ppm Cr along the magmatic front to 60–80 ppm Ni and ~140 ppm Cr for W. Guguan and 100–380 ppm Ni and 260–900 ppm Cr for Guguan 2.

Variations in alkali metal contents indicate that magmatic front melts reflect higher degrees of melting than those to the rear. Na6 (defined as Na2O contents at 6% MgO) varies inversely with melt fraction (F) globally (Plank and Langmuir 1988). Guguan and N. Guguan lavas together define a trend indicating Na6~1.9%, at the low end of the range of Na6 for Mariana volcanoes (Maug 1.9%, Pagan 2.5% and Sarigan 2.7%; Plank and Langmuir 1988), but similar to the mean for Izu arc volcanoes (Na6=1.9±0.3%). Using the Na--Mg regression for Guguan and N. Guguan lavas, basalts from W. Guguan (Na6~2.8%) and Guguan 2 (Na6~3.3%) represent significantly lower degree melts than those from beneath the magmatic front. The covariation of alkali metals is noteworthy; the increase in Na6 of ~70% from the magmatic front to Guguan 2 is similar to the increase in K55 of ~50% and suggests that these variations are due to similar controls, probably F. Note that Guguan 2 is associated with a potential melting column that is about twice as high as that beneath Guguan but erupts melts formed by a smaller extent of melting (lower F). This indicates that variations in F beneath cross-chains are not controlled by the height of the melting column, a variable which globally explains most of the differences found for F associated with the magmatic front of arcs (Plank and Langmuir 1988).

It is important to note that there is no significant variation in water contents in cross-chain lavas, as evidenced by olivine-hosted GI (Fig. 3c; A. Shaw, written communication, 2004). Olivines in one Guguan sample (GUG-3) have mean H2O (n=75) of 2.64±0.43% (1 SD), similar to a mean H2O (n=49) of 2.54±0.52% in olivine-hosted GI in two N. Guguan (D50) samples. This estimate of ~2.6% H2O is within the range of 2–4 wt% H2O in Guguan GI (Kelley 2002) and 1.4--3.3% H2O in Mariana arc basaltic GI determined by Newman et al. (2000). Paradoxically, GI from olivines in Guguan 2 lavas are not drier than melts from the magmatic front; GI in D46 olivines have mean H2O (n=19) of 2.36±0.71%, whereas those from D48 have mean H2O (n=6) of 3.74±1.86%.

Trace element variations

Trace element variation diagrams (Table 3, Fig. 4) show that all Guguan cross-chain lavas are enriched in fluid-mobile incompatible elements (Rb, Ba, U, K, Sr, and Pb) and depleted in high field strength incompatible elements (HFSE: Nb, Zr, Ta, HREE), as is typical of convergent margin magmatic suites. Rare Earth Element (REE) abundances vary systematically along the cross-chain. Chondrite-normalized ratios (X/Y) n of REE reveal that magmatic front lavas have flat REE patterns, (mean (La/Nd) n  = 0.95 for Guguan, 1.06 for N. Guguan) but become increasingly LREE-enriched (mean (La/Nd) n  = 1.18 and 1.39 for W. Guguan and Guguan 2, respectively) rearword, similar to rear-arc enrichments shown by the K--h relationship. HREE patterns are also nearly flat for Guguan and N. Guguan ((Gd/Yb) n  = 1.12 and 1.06, respectively), and are slightly steeper for W. Guguan ((Gd/Yb) n  = 1.28) and Guguan 2 ((Gd/Yb) n  = 1.30). Europium anomalies are trivial, with Eu/Eu*=0.99 and 1.02 for Guguan and N. Guguan, respectively, 0.98 for W. Guguan, and 0.97 for Guguan 2. There are modestly negative Ce anomalies for Guguan (Ce/Ce*~0.94) but not for N. Guguan (Ce/Ce*=1.02), W. Guguan (0.98), or Guguan 2 (0.97).

Table 3 Trace element data
Full size table
Fig. 4
figure 4

Trace element variation diagram, arranged in order of elements becoming increasingly incompatible in the mantle to the left. Means for each edifice are shown. Data for Guguan are from Elliott et al. (1997), Woodhead (1989), and Woodhead et al. (2001). Element sequence and normalizing values after Hofmann (1988). MTSB, Mariana Trough Basalt formed by seafloor spreading

Full size image

Some ratios of fluid-mobile incompatible elements (such as Rb, U, Ba, Sr, and Pb) to HFSE and REE (which are generally not thought to be transported in hydrous fluids, but see different conclusions of Woodhead et al. (2001)) vary as a function of height above the WBZ, but others do not. Ba/Th and U/Th in Guguan cross-chain lavas are compared in Fig. 5 to typical ratios for MORB (Hofmann 1988) and for bulk sediments about to be subducted beneath the Marianas (Plank and Langmuir 1998). U/Th and Ba/Th inversely vary with depth to the WBZ (Fig. 5). U/Th decreases from ~0.5 on Guguan and ~0.7 on N. Guguan to ~0.48 on W. Guguan and reaches mantle-like values (~0.38) for Guguan 2 lavas (MORB U/Th ~0.38, subducted sediment ~0.22). Ba/Th is taken by Pearce et al. (2005) to track the subduction component carried in hydrous fluids; this decreases from very high values along the magmatic front (Guguan range: 300–700; mean = 442, although N. Guguan Ba/Th is much lower: 140–210), to lower values for W. Guguan (~210) and Guguan 2 basalts (130–190). All of these lavas show significantly higher Ba/Th than ~75 typical of MORB (Hofmann 1988) or sediment subducted beneath the Marianas (~120; Plank and Langmuir 1998).

Fig. 5
figure 5

Behavior of diagnostic incompatible trace elements in mafic Guguan cross-chain lavas. a Ba/Th vs. La/Sm, modified after Elliott (2003); b U/Th vs. Zr/Nb. In both cases, the element pair with the fluid-mobile element in the numerator is plotted along the y-axis, whereas the x-axis plots ratios that are sensitive to source depletion. Composition of MORB from Hofmann (1988), composition of IBM sediment melt from Hochstaedter et al. (2001)

Full size image

Pearce et al. (2005) argue that Ba/Ta (and Ba/Nb) is a good proxy for the total `subuduction component’ addition to the arc magma source. Ba/Nb decreases from ~120–200 for Guguan and N. Guguan to 70–80 for W. Guguan and 50–100 for Guguan 2 basalts, still much higher than Ba/Nb ~4 for N-MORB (Hofmann 1988) or ~19 for sediment subducted beneath the Mariana arc (Plank and Langmuir 1998). The depth of the Nb anomaly, as measured by La/Nb (patterns with no anomaly have La/Nb ~1.1), does not change significantly across the cross-chain, varying from 3.5 to 6.0.

These relationships indicate that the relative contribution of fluidmobile elements U and Ba decrease with increasing slab depth. Other incompatible element pairs with a fluid-mobile element (Rb, Sr, Pb) in the numerator do not show such systematic variations. This includes Sr/Nd (~38 on Guguan and ~73 on N. Guguan; ~39 on W. Guguan and ~38 on Guguan 2), Pb/Ce (~0.23 for Guguan and ~0.37 for N. Guguan; 0.30 for W. Guguan, 0.24 for Guguan 2), and Rb/Zr (~0.15 for Guguan, ~0.17 for N. Guguan, ~0.14 for W. Guguan, ~0.2 for Guguan 2). Ba/Zr also shows little variation across the cross-chain (~2.9 for Guguan and ~3.1 for N. Guguan; 2.3 for W. Guguan and ~3 for Guguan 2). These ratios are nonetheless higher than fresh MORB (Sr/Nd~15, Pb/Ce~0.041, Rb/Zr~0.012, and Ba/Zr~0.133 (Hofmann 1988). Relative to ratios in bulk sediments about to be subducted beneath the Marianas (Sr/Nd~7.7, Pb/Ce~0.19, Rb/Zr~0.35, and Ba/Zr~3.6 (Plank and Langmuir 1998)), cross-chain lavas have much higher Sr/Nd, similar Pb/Ce and Ba/Zr and lower Rb/Zr. These variations support the conclusion that a ‘subduction component’ is carried in a hydrous fluid along the magmatic front---at least for Guguan itself---and that the contribution of this component decreases variably as WBZ increases.

Incompatible element ratios that monitor source depletion vary systematically across the Guguan cross-chain (Fig. 5). La/Yb and Zr/Y should decrease as melting proceeds, whereas Zr/Nb increases with source depletion. La/Yb increases from ~1.7 for Guguan and ~1.6 for N. Guguan (values slightly greater than La/Yb ~1.0 typical of N-MORB) to 2.4 for W. Guguan and ~3.5 for Guguan 2. Zr/Y increases slightly, from ~2.47 for Guguan and ~1.8 for N. Guguan to ~2.6 for W. Guguan and ~2.6 for Guguan 2 (MORB value ~2.9). Zr/Nb is high along the magmatic front (~60 for Guguan, ~70 for N. Guguan), decreasing rearwards (~39 for W. Guguan, ~34 for Guguan 2), approaching values for MORB (~30). Nb/Yb in melts is expected to decrease as the mantle source becomes increasingly depleted (Pearce et al. 2005). Nb/Yb is at a minimum along the magmatic front and increases from ~0.4 for Guguan and N. Guguan to ~0.8 for W. Guguan and ~1 for Guguan 2 basalts, values that approach MORB Nb/Yb ~0.92 (Hofmann 1988). These relations indicate that the mantle source of the rear-arc magmas is less depleted than that supplying the magmatic front, but cannot be due only to progressive depletion. La must be added and/or Nb subtracted from the mantle source region.

The Ba/Th vs. La/Sm diagram (Fig. 5a) was suggested by Elliott (2003) to distinguish two components responsible for much of the compositional variation observed in subduction zone magmas. The high Ba/Th, low La/Sm endmember represents melts from mantle modified by fluids from altered oceanic crust, an interpretation that is consistent with other lines of evidence for Guguan lavas, which plot closest to this endmember. The significance of the low Ba/Th, high La/Sm endmember is ascribed to sediment melt (Elliott 2003). This endmember could also manifest participation of enriched OIB-type mantle (Lin et al. 1989). Guguan 2 lavas approach the low Ba/Th, high La/Sm endmember. It is difficult to explain this as a manifestation of sediment melting because Th/Nb, also taken as an index of sediment melting by Elliott et al. (1997), does not vary systematically across the cross-chain. Th/Nb ranges from ~0.43 for Guguan and ~1.0 for N. Guguan to ~0.42 for W. Guguan and ~0.50 for Guguan 2. This is an order of magnitude higher than Th/Nb ~0.05 for MORB and is also significantly higher than Th/Nb ~0.24 for subducted sediment. It is similar to the value of 0.7 inferred for sediment melts beneath the northern IBM arc (Hochstaedter et al. 2001); however, the modest variations observed for the cross-chain do not indicate a more important role for sediment melts in rear-arc volcanoes relative to the role that this component plays in magmatic front melt generation.

The U/Th vs. Zr/Nb diagram (Fig. 5b) provides a different perspective on the origin of geochemical variations for the Guguan cross-chain. High U/Th is expected for mantle sources that have been metasomatized by hydrous, slab-derived fluids, whereas high Zr/Nb is expected for depleted mantle sources. Magmatic front lavas have higher U/Th and Zr/Nb than do those of rear-arc lavas, suggesting linkages between source depletion and metasomatism due to hydrous, slab-derived fluids. In contrast Guguan 2 lavas have lower U/Th and Zr/Nb and cluster near MORB, OIB (not shown on Fig. 5b), and sediment.

Isotopic variations

Isotopic compositions of Sr, Nd, Hf, and Pb (Table 4) vary with height above the WBZ. Sr and Nd vary the most, whereas Hf and Pb do not vary as systematically. 87Sr/86Sr decreases from a mean of 0.70347 for Guguan and 0.70350 for N. Guguan to 0.70320 for W. Guguan and 0.70317 for Guguan 2 (Fig. 6). We infer from the rearward decrease in 87Sr/86Sr that the mantle beneath the magmatic front is more perfused with radiogenic Sr extracted from sediments and altered oceanic crust than is the mantle farther from the trench. Nd isotopic compositions (143Nd/144Nd, reported as εNd) also decrease across the cross-chain, from a mean of +7.8 for Guguan and +8.1 for N. Guguan to +7.4 for W. Guguan and +6.9 for Guguan 2, although scatter of ~2 epsilon units is observed for each edifice (Fig. 6).

Table 4 Isotopic data
Full size table
Fig. 6
figure 6

Plot of Nd and Sr isotopic composition of Guguan cross-chain lavas. Cross-chain data are compared with lavas from the Mariana island are (grey field, Stern et al. 2003) and with data for Philippine Sea Plate (PSP) ‘basin’ lavas, including samples from the W. Philippine Basin and Parece Vela-Shikoku Basin (Hickey-Vargas 1998). Mixing curves are shown for mixtures between depleted MORB-type mantle (DMM) and sediments on the subducting Pacific plate. Percent mixtures of sediment (after Plank and Langmuir 1998) are also shown. Note that the isotopic composition of Guguan 2 and W. Guguan lavas falls in the field defined by basalts from PSP basins (Hickey-Vargas 1998)

Full size image

The observed covariation of Sr--Nd isotopic compositions for Guguan cross-chain lavas is opposite to that observed for Mariana and other arc--back-arc basin magmatic pairs, for which BABB usually have higher 143Nd/144Nd and lower 87Sr/86Sr than associated arc lavas. In the case of the Marianas, εNd increases from a mean of +6.7 for lavas from the IBM magmatic front to a mean of +8.9 for Mariana Trough BABB, whereas mean 87Sr/86Sr decreases from the arc (0.70352) to Mariana Trough BABB (mean 87Sr/86Sr= 0.70291; Stern et al. 2003). The isotopic covariation seen for the Guguan cross-chain is also perpendicular to the mantle trend, where the first-order feature is the antivariation of Sr and Nd isotopic compositions. Covariation of Sr and Nd isotopic compositions is characteristic of other arc cross-chains (Hochstaedter et al. 2001; Ishizuka, et al. 2003; Stern et al. 1993; Tatsumi 2003; Woodhead et al. 1998). The antivariation of Sr and Nd isotopic compositions observed across the Kasuga cross-chain (Fig. 1b) led Stern et al. (1993) to identify a ‘Frontal Arc Trend’ (FAT) because lavas from the magmatic front are readily distinguished from rear-arc lavas by having higher 87Sr/86Sr at a given143Nd/144Nd, whereas rear-arc lavas define a lower 87Sr/86Sr--143Nd/144Nd ‘Rear Arc Trend’ (RAT). The Guguan cross-chain isotopic data also define a FAT--RAT trend that is consistent with two hypotheses. Both hypotheses accept that the FAT trend manifests addition of subducted, radiogenic Sr, largely delivered by aqueous fluids to the source of melt generation. The first hypothesis holds that the RAT trend reflects normal mantle, relatively unmodified by additions from subducted slab and sediments, whereas the second hypothesis holds that the RAT trend reflects increasing participation of sediment melts.

Pb isotopic compositions of Guguan cross-chain samples lie in the field for Mariana arc lavas (Fig. 7). Guguan lavas range in 206Pb/204Pb from ~18.5 to ~18.8, whereas data for N. Guguan (~18.75) and W. Guguan (~18.7) cluster tightly. The two dredges from Guguan 2 vary considerably, with D46 being significantly less radiogenic (206Pb/204Pb~18.7) than D48 (206Pb/204Pb~18.85). All 208Pb/204Pb data for the Guguan cross-chain plot along the Northern Hemisphere Reference Line (NHRL) and show no evidence of elevated 208Pb/204Pb characteristic of some Mariana Trough basalts as well as older lavas of the Parece Vela-Shikoku and W. Philippine Sea basins to the west (Hickey-Vargas 1998). On Pb isotope plots, Guguan cross-chain samples---like all analyses of the active Mariana arc---lie in the Pacific (as opposed to Indian ocean) field as defined by (Pearce, et al. 1999). In common with other Mariana arc lavas, Guguan cross-chain lavas have 207Pb/204Pb that is slightly to significantly displaced from NHRL towards the field for sediments. Similar observations have led others to conclude that the Pb in Guguan cross-chain lavas are derived from Pacific-type subducted sediments and crust (Woodhead and Fraser 1985), but it has not been explained why 208Pb/204Pb is not similarly displaced towards the field of sediments.

Fig. 7
figure 7

Pb isotopic composition of Guguan cross-chain lavas. Fields for subducting sediments, Mariana Trough basalts, Mariana arc lavas, and Cretaceous off-ridge volcanics from Stern et al. (2003). NHRL, Northern Hemisphere Reference Line (Hart 1984). Fields for Mariana Trough (thin solid line) and Mariana arc (thick solid line) as well as units about to be subducted: sediments (grey) and Cretaceous off-ridge volcanics (dashed line) are from Stern et al. (2003). Bold “V” marks composition of OIB volcaniclastics (Elliott et al. 1997; Woodhead and Fraser 1985)

Full size image

Seafloor subducted beneath Guguan should include volcaniclastic sediments and lava flows produced by off-ridge volcanic activity in mid-Cretaceous time. A possible explanation for the Sr--Nd RAT trend is that covariations in Sr and Nd isotopic compositions in Guguan cross-chain lavas are sourced from subducted OIB volcaniclastics and lavas, some of which have HIMU affinities. The Pb isotopic data do permit the possibility that OIB-related volcaniclastics (`V’ on Fig. 7) could be important sources of Pb in Guguan cross-chain lavas. In particular, Guguan 2 D48 samples have slightly elevated 206Pb/204Pb relative to other cross-chain lavas, and this could be due to a significant contribution from HIMU volcaniclastic sediments. This possibility is inconsistent with the observation that D46 lavas plot closer to HIMU than do D48 lavas on the Sr--Nd isotopic diagram (Fig. 6).

Hf isotopic data plot in the field of Indian Ocean basalts (Fig. 8a) as defined by Pearce et al. (1999). There is relatively little variation in εHf observed across the cross-chain (mean for Guguan=15.9; N. Guguan=14.9; W. Guguan=16.2; Gugan 2=15.0), but the data suggest a trend that lies perpendicular to that for mixing of asthenospheric mantle and sediments (Fig. 8b).

Fig. 8
figure 8

a Hf--Nd isotopic data for Guguan cross-chain samples. Dividing line between Indian and Pacific mantle sources is from Pearce et al. (1999). Note that all Guguan cross-chain samples, like all Mariana arc samples, plot in the Indian Ocean mantle field. Fields for global MORB and OIB are from Vervoort et al. (1999). b Hf--Sr isotopic data for Guguan cross-chain samples. Curve shows mixing trajectory for sediment mantle mixtures, with % sediment shown. Sediment isotopic compositions are from Plank and Langmuir (1998) for Sr and J.D. Woodhead (personal communication) for Hf

Full size image

Discussion

It is generally acknowledged that most convergent margin melts---especially those where old (>40 Ma), cold lithosphere is subducted---are generated because the flux of hydrous fluids from the subducted slab causes mantle to melt at a much lower temperature than is normal (Gaetani and Grove 2003). Geochemical and isotopic data for Guguan cross-chain lavas, coupled with existing data for Guguan, provide a valuable perspective on several aspects of how the flux of fluids and cations from the subducted slab as well as how magma generation processes in the mantle vary with depth to the WBZ. A modest K--h relationship holds for this system, and this is because the integrated degree of melting increases towards the magmatic front. It is not clear if this reflects differences in the proportion of melting beneath each of the cross-chain volcanoes or if this is cumulative, the additive effects of sequentially melting the mantle that is advected from the rear arc towards the magmatic front. To address this question, we first examine the possible role of multiple episodes of melting resulting from the interaction of induced convection and fluxing from the subducted slab. We turn to the issue of whether or not cross-chain lava compositions indicate a progressively greater role for sediment melting with increasing depth to the WBZ.

Sequential melting as a result of induced convection

Several investigators (McCulloch and Gamble 1991; Woodhead et al. 1993) infer that the overriding asthenospheric mantle wedge is sequentially melted as it overturns beneath a convergent plate margin. In this scenario, mantle rises beneath the back-arc basin spreading ridge and migrates towards the trench before being dragged down by the descending plate. Multiple opportunities for melting exist in this scenario, first pressure-release melting beneath the back-arc basin, followed by flux melting as the convecting asthenosphere approaches the magmatic arc, first beneath the cross-chains and finally beneath the magmatic front (Iwamori 1998). Melting beneath the arc may also be partly caused by decompression melting (Conder et al. 2002; England et al. 2004). Such a scenario for sequential melting beneath the Mariana Trough back-arc basin and Guguan cross-chain, modified after models of McCulloch and Gamble (1991) and Woodhead et al. (1993) is shown in Fig. 9a.

Fig. 9
figure 9

Sequential melting in arc--backarc basin systems. a Conceptual model (no vertical exaggeration) for sequential melting beneath the Guguan cross-chain, modified after Stern et al. (2003). Note the multiple melting opportunities for convecting asthenosphere: beneath the spreading axis (1), beneath rear-arc cross-chain volcanoes (2), and beneath the magmatic front (3). b Nb/Yb vs. Yb, and c Zr/Y vs. Y systematics for Guguan cross-chain and Mariana Trough basalts (MTB); samples with <54% SiO2 only are plotted. Trajectories show batch nonmodal melting of spinel peridotite. Model parameters are modified after Woodhead et al. (1993). The thin line corresponds to batch melting of ‘primitive mantle’ (Hofmann 1988; light grey box ‘PM’); the thick line corresponds to melting of the residue (RSD, dark grey box) left after 10% batch melting of primitive mantle. Data for the Mariana Trough are from Pearce et al. (2005) and other published sources

Full size image

This model can be examined using simple models of batch melting. Fig. 9b and c plot ratios for ‘conservative’ (in the sense of Pearce and Peate 1995) incompatible elements (Nb, Yb, Zr, and Y), which are not thought to be mobile in fluids derived from the subducting slab. Figure 9b and c also shows trajectories for melting of ‘primitive mantle’ and the expected composition of mantle residue after 10% batch melting. For both Nb--Yb and Zr--Y systematics, Mariana Trough basalts approximate trajectories expected for batch melting of primitive mantle, although these basalts have systematically lower Nb contents than simple models predict. This is not surprising in view of the fact that BABB, like all convergent margin magmas, typically are relatively depleted in Nb relative to MORB or OIB. Ranges of ~5--15% partial melting seen on Fig. 9b and c are consistent with estimates of 13±5% melting for Mariana Trough basalts that formed in association with seafloor spreading (Gribble et al. 1998).

Lavas erupted along the magmatic front of arcs characteristically show chemical signatures (Na2O, HFSE concentrations) indicating significantly higher degrees of melting than basalts from any other tectonic environment. Arc magmas result from 10–30% melting, with lithospheric thickness and water content being the most important controls (Pearce and Peate 1995). The relative importance of these two factors can be assessed using variations in the extent of melting inferred for arc cross-chains. Rear-arc volcanoes are associated with a potentially much longer melting column than is the case for magmatic front volcanoes because they lie over much deeper parts of the WBZ, and in the case of the Guguan cross-chain are associated with thinner lithosphere. Nevertheless, melts from theses rear-arc volcanoes manifest lower degrees of melting, as discussed previously. Variations observed for HFS trace elements (Fig. 9b, c) are difficult to interpret as unequivocally indicating systematic differences in melting across the cross-chain. We are, nevertheless, impressed by the fact that the rear-arc lavas do not appear to be derived from a mantle source that was strongly depleted by melting beneath a back-arc basin spreading center. Guguan 2 in particular appears from trace element and isotopic perspectives to be derived from relatively undepleted mantle. This is only one cross-chain, but it is the only one that we know to be associated with an actively spreading back-arc basin in an IOA setting. The story it tells is that the mantle source of rear-arc volcanoes tap a relatively undepleted mantle source.

It appears from trace element perspectives that Guguan and N. Guguan melts were derived from a more depleted source than were those of W. Guguan and Guguan 2 (Fig. 5b). This suggests that a model of sequential melting may still be applicable to progressive depletion of cross-chain sources only, from a source that is not affected by melt depletion in the back-arc basin. A similar point has been made for Central America (Cameron et al. 2002) and Kamchatka (Hochstaedter et al. 1996) arcs. An alternate explanation is that the Guguan 2 source was depleted by melting beneath the Mariana Trough spreading ridge but has been overprinted by sediment melts so that it appears undepleted. We will show in the next section that this is not likely, because trace element and isotopic data are inconsistent with the addition of sediment melts from the subducted slab.

Another possibility is that the Guguan cross-chain represents an example of ‘hot fingers’ in the mantle (Tamura et al. 2002). We note that the shoalest part of the Mariana Trough back-arc basin lies about 17°N, WSW of the Guguan cross-chain (Martinez and Taylor 2003), suggesting a greater-than-normal extent of melting for this region. This part of the Mariana Trough and the Guguan cross-chain may be surficial expressions of a convecting mantle thermal anomaly. Evaluating this possibility requires shallow mantle tomographic images and thus is beyond the scope of this paper.

Was the mantle source of Guguan cross-chain magmas affected by sediment melts?

Several workers have noted the relatively enriched nature of cross-chain lavas and argued that these enrichments result from mixing sediment melt into the asthenospheric mantle source region (Hochstaedter et al. 2000; Ishizuka et al. 2003). This is an attractive explanation for the K--h relationship, because experimental results indicate that sediment melting is favored by the higher temperatures found at greater depths in the WBZ (Johnson and Plank 1999), although it should be noted that distinctions between ‘fluid’ and ‘melt’ at pressures >6 GPa (depths >180 km) do not exist (Kessel et al. 2005). Sediment melting also explains many aspects of incompatible trace element behavior for rear-arc lavas, such as Ba/La vs. La/Yb and U/Th vs. Zr/Nb (Fig. 5), where Guguan 2 lavas plot close to sediments or sediment melts. The argument for sediment melts is more difficult to reconcile with isotope data. If enrichments in Sr and Nd isotopic compositions found for rear-arc lavas are due to an increased contribution of sediments, it is reasonable to expect that these lavas should lie along an isotopic array that extends from depleted mantle towards subducted sediment. Figure 6 shows two such mixing trajectories, and Guguan cross-chain lavas trend perpendicular to both of these.

The Guguan cross-chain defines a trend that is perpendicular to that expected for mixing between MORB-type asthenospheric mantle and sediments, and this typifies global IOA cross-chain isotopic systematics. Figure 10 plots Sr and Nd isotopic compositions for several IOA cross-chains. Note that with the exception of the Kasuga cross-chain, 87Sr/86Sr generally decreases with increasing depth to the subducted slab, opposite to what would be expected if sediment melting became increasingly important with depth. Nd isotopic compositions vary less systematically as a function of depth to the subducted slab, especially considering New Britain and Kasugas, but for Guguan and the Izu cross-chains, Nd isotopic compositions also decrease with increasing WBZ depth. Thus, intra-oceanic cross-chains, considered together or individually, show little isotopic evidence for the participation of sediment melts in their source. Indeed, the W. Guguan and Guguan 2 plot in the field defined by basalts of the Philippine Sea plate, for which little subduction component has been identified (Hickey-Vargas 1998).

Fig. 10
figure 10

a Plot of 87Sr/86Sr vs. height above the subducted slab; and b 143Nd/144Nd for five intra-oceanic cross-chains. Data for Guguan from this study, Central Izu from Hochstaedter et al. (2001) and Taylor and Nesbitt (1998), N. Izu from Taylor and Nesbitt (1998) and T. Yoshida (personal communication), Kasuga from Stern et al. (1993), New Britain from Woodhead et al. (1998) and Woodhead and Johnson (1993)

Full size image

Pb isotopic data for Guguan cross-chain lavas are elevated towards the field for subducted sediments in 207Pb/204Pb vs. 206Pb/204Pb but not in 208Pb/204Pb vs. 206Pb/204Pb, although the Pb isotopic trends do extend towards OIB volcaniclastics (Fig. 7). With the exception of more radiogenic D46 samples, there is no systematic difference between the Pb isotopic composition of lavas from the volcanic front and those from rear-arc volcanoes. In particular, cross-chain lavas do not plot closer to the field for subducted sediments than do lavas from along the magmatic front. Although not shown, it is worth noting that the elevated 208Pb/204Pb characteristic of Izu cross-chain lavas is not observed for the Guguan cross-chain. This suggests that an important isotopic province boundary lies somewhere between these beneath the IBM arc.

As was noted for Sr--Nd isotopic systematics, Sr--Hf isotopic variations plot along a trend perpendicular to the MORB mantle--sediment mixing trajectory. This is inconsistent with participation of sediment melts in the formation of rear-arc magmas, for the same reasons as was the case for Sr--Nd isotopic systematics.

Finally, Fig. 11 plots Th/Nb, taken as an index of sediment melt contribution (Elliott et al. 1997), versus Sr isotopic composition. Once again, the array defined by Guguan cross-chain samples is orthogonal to a mixing trajectory between MORB mantle and sediment or sediment melt. Elliott et al. (1997) argue that sediment melts in equilibrium with residual rutile could have much higher Th/Nb than the sediment itself, in which case the mixing curve could have a lower angle. Given that large variations in Th/Nb are observed for Guguan cross-chain lavas without a significant change in Sr isotopic composition, it is very difficult to endorse the idea that Th/Nb variations reflect sediment melt contributions. A similar lack of systematic variation between Th/Nb and 87Sr/86Sr led Hochstaedter et al. (2001) to disavow a role for sediment melting in controlling the K--h relationship of Izu arc cross-chains. In conclusion, we cannot find any convincing evidence that melts of subducted sediments are important for controlling the isotopic compositions of Sr, Nd, Pb, and Hf in Guguan cross-chain lavas.

Fig. 11
figure 11

Plot of 87Sr/86Sr vs. Th/Nb for Guguan cross-chain samples. Symbols as in Fig. 8. Mixing trajectory between asthenospheric depleted mantle (DM) and Western Pacific sediment (Plank and Langmuir 1998) is shown. Approximate composition of sediment melt is from Hochstaedter et al. (2001); this endmember would have higher Th/Nb and more radiogenic Sr than unmelted sediment, but data for Guguan cross-chain would still be sub-perpendicular to mixing trajectories between sediment melt and DM. Grey field denotes the composition of lavas from the Central Izu cross-chains (Hochstaedter et al. 2001; Taylor and Nesbitt 1998). Also shown is composition of Mariana Trough Spread basalt (MTSB )

Full size image

Another possibility is that incompatible elements in the subducted sediments are strongly fractionated by devolatilization at shallower depths, and that melts or fluids derived from such devolatilized sediments are responsible for the enrichments observed in the cross-chain seamounts. Ryan et al. (1996) stressed that subducted slabs progressively devolatilize during descent, but also stressed the difficulty of distinguishing such fractionations due to this process versus those in the mantle. Hochstaedter et al. (2001) argued that the subduction component for rear-arc volcanoes in central Izu cross-chain was a fluid dehydrated from a residual slab that had already been depleted in fluid-mobile elements beneath the volcanic front. In their model, the ‘depleted’ fluid is added to elementally and isotopically more enriched mantle beneath rear-arc volcanoes. A similar model may serve to explain the chemical composition of Guguan cross-chain lavas but not the isotopic data.

A greater role for water from Serpentinite with greater depth to the WBZ?

The variations in mineral chemical, geochemical, and isotopic compositions observed across the Guguan cross-chain are consistent with a reduced subduction component rearward. This is shown by the fact that some ratios of fluid-mobile to fluid-immobile trace elements and Sr--Nd isotopic variations in reararc lavas are distinct from magmatic front lavas and similar to OIB and MORB. These relations suggest that Guguan cross-chain rear-arc melts should be relatively anhydrous, but as Fig. 3a shows, GI in these lavas contain at least as much water as those from the magmatic front. Correcting for fractionation makes the frontal arc GI less even drier. This results in a paradox, whereby rear-arc magmas contain as much (or more) water as magmas from the magmatic front but show less trace element or isotopic evidence of a subduction component.

An intriguing possibility is that the sources of magmatic water---and the principal metasomatic agent---changes with greater WBZ depth from ones that also contain readily mobilized LIL elements, such as sediments or altered MORB, to sources that yield abundant water but little fluid-mobile large-ion lithophile elements (LILEs). Serpentinite is the best candidate, because it is extremely LIL-depleted and can carry water to greater WBZ depths than can either altered MORB or sediments (Schmidt and Poli 1998). Serpentinized mantle contains ~12% H2O (Schmidt and Poli 1998) and this water, when released by dehydration, would carry little LILE. Faulting at the outer trench swell may allow water to seep deep into the cold lithosphere about to be subducted, serpentinizing much of the shallow mantle lithosphere (Ranero et al. 2003). In addition, sub-forearc mantle is pervasively serpentinized by fluids emanating from the downgoing slab and sediments (Peacock and Hyndman 1999), and much of this serpentinite may be entrained as part of a `viscous blanket’ developed adjacent to the downgoing slab (Kincaid and Sacks 1997). Both of these serpentinite zones may yield water upon dehydration (Yamasaki and Seno 2003). Most cold subduction zones---including the Marianas---have two parallel zones of seismicity, and the lower zone appears to manifest where serpentinite on the downgoing slab is dehydrating (Peacock 2001).

Rüpke, et al. (2004) infer from numerical models that water is continuously released from the subducting slab with the host reservoir changing with depth. They modeled the dehydration of ~120 Ma plate, and these results indicate that sediment generally dehydrates first (at depths of ~50 km), then oceanic crust (100–200 km depth), and finally serpentinized mantle (~150–250 km depth). Guguan 2 lies ~230 km above the WBZ, allowing the possibility that magmatic water for these lavas is largely produced by dehydrating serpentinite.

The Guguan cross-chain data do not allow us to confidently interpret the geochemical and isotopic variations to be due to a change in the sources of water and LILEs in subducted oceanic crust beneath Guguan to those in subducted serpentinite and enriched mantle beneath Guguan 2. Testing this hypothesis requires determining magmatic water as well as abundances and isotopic compositions of elements that should be released from dehydrating serpentinite, such as B and Cl.

Conclusions

The Guguan cross-chain is an excellent example of convergent margin magmatism where mafic melts produced from a range of depths above the WBZ can be examined without complications arising from interactions with continental crust or lithosphere. Guguan cross-chain basalts were produced at depths of 125 km beneath the magmatic front to 230 km and show systematic variations with increasing depth, including: (1) decreasing fractionation; (2) lower degrees of melting, in spite of being associated with much higher potential melting columns and similar water contents; (3) coexisting olivine—plagioclase--spinel compositions become more like those found in MORB and OIB; (5) a modest K--h relationship; (6) involvement of less-depleted mantle; (7) relative trace element abundances indicate strong gradients in some fluid-mediated subduction components, especially U/Th and Ba/La; (8) other trace-element ratios thought to monitor fluid-mediated subduction component do not decrease with increasing WBZ depth, including Sr/Nd, Rb/Zr, Ba/Zr, and Pb/Ce, nor does the magnitude of the Nb anomaly change much; (9) isotopic compositions of Sr, Nd, and Hf become less radiogenic; (10) Pb isotopic compositions do not change significantly. These results are consistent with a model of sequential melting of mantle beneath the cross-chain. The observed variations are difficult to reconcile with the hypothesis that rear-arc lavas manifest a subduction component carried in sediment melt, and are more simply explained as a result of a reduced subduction component, perhaps released by dehydrating subducted serpentinite, being added to relatively enriched mantle.