The Quaternary Southwest Japan Arc is a product of subduction of the hot, young Philippine Sea Plate beneath the Eurasian Continental Plate. The magmas erupted from the Southwest Japan Arc belong to a category of magmas commonly referred to as “adakites” or “adakitic magmas”. These magmas show trace element evidence for interaction with garnet at depth, and may be associated with partial melting of subducted altered oceanic crust. Also found throughout the southern Sea of Japan region are alkali basalts with little apparent connection to the subduction zone. We have determined major element, trace element, and Sr, Nd, Pb, and U-Th isotopic compositions for a bimodal suite of lavas erupted at the Daisen volcanic field in the Southwest Japan Arc. These magmas consist of mildly alkaline basalts and a calcalkaline intermediate suite, separated by a ∼10 wt.% silica gap. The intermediate magmas show trace element and isotopic evidence for interaction with garnet, consistent with partial melting of the hot, young (∼20 Ma) Philippine Sea Plate. The Daisen intermediate magmas are distinct from other adakitic magmas in their radiogenic isotopic characteristics, consistent with a significant contribution (∼25%) from subducted Nankai Trough sediments. Our data suggest that the basalts erupted at the Daisen volcanic field are not parental to the intermediate magmas, and contain a small contribution of EM1-like mantle common in Sea of Japan alkali basalts but not apparent in the Daisen intermediate magmas.
 Since the Archean, a particular subclass of arc-associated igneous rocks displaying geochemical characteristics indicating equilibration with garnet has been generated in locations worldwide. These rocks are characterized by high Sr/Y, steeply fractionated middle-to-heavy rare earth elements (REE; e.g., high Dy/Yb), and, in some younger rocks, (238U/230Th) < 1 (Th activity excess). These have commonly been referred to as “adakites” or “adakitic magmas” [Defant and Drummond, 1990], in honor of their first description in samples from Adak Island in Alaska [Kay, 1978]. First proposed to be partial melts of exceptionally warm subducted oceanic crust [Kay, 1978; Defant and Drummond, 1990; Drummond and Defant, 1990], it has since become clear that there is a variety of circumstances under which arc magmas might equilibrate with garnet. For example, garnet is expected to be stable during partial melting of underplated basaltic materials at the base of thick continental crust [Kay et al., 1987; Atherton and Petford, 1993]. Crustal materials might be drawn into the mantle wedge, stabilizing garnet, by the process of forearc subduction erosion [Kay et al., 2005; Goss and Kay, 2006]. And although direct evidence in the form of phenocrysts or cumulate xenoliths is rare, experiments suggest that garnet can be a stable liquidus phase in arc magmas under certain conditions, allowing for direct fractional crystallization of garnet [Alonso-Perez et al., 2009; Coldwell et al., 2011].
 The Southwest Japan Arc is formed by the subduction of the Philippine Sea Plate beneath the Eurasian Continental Plate (Figure 1 inset). The Philippine Sea Plate is young (presumably equivalent in age to the Shikoku Basin, ∼17–25 Ma [Hickey-Vargas, 1998]) and warm where it subducts beneath the Southwest Japan Arc, in contrast to the typical old and cold Pacific Plate that is subducted beneath the Northeast Japan Arc, resulting in a thermal regime that is >300° hotter beneath the volcanic front in the Southwest compared to the Northeast [Peacock and Wang, 1999]. The young, warm slab makes the Southwest Japan Arc an ideal site for slab melting, and previous authors have shown that the Quaternary arc lavas in Southwest Japan display the “adakitic” geochemical characteristics described above, while those in Northeast Japan do not [e.g., Morris, 1995]. The distinctive trace element characteristics of the Southwest Japan Arc lavas (i.e., high Sr/Y, steeply fractionated REEs) have been interpreted variously to be the result of garnet fractionation during slab melting [Morris, 1995], equilibration with garnet-bearing mantle at depths of ∼60 km [Tamura et al., 2003], and fractional crystallization of garnet from relatively dry parental magmas at the base of the arc crust [Zellmer et al., 2012]. The goal of this study is to assess the relative contributions of various inputs including mantle sources, Philippine Sea crust, subducted sediments, and lower crustal materials to magmas associated with the Daisen volcanic field in Southwest Japan. We also aim to clarify the relationship between the basalts erupted at the western margin of the Daisen volcanic field and the intermediate lavas that comprise the bulk of the edifice.
2. Geologic setting of Daisen Volcanic Field
 The Daisen volcanic field is located in the Quaternary Southwest Japan Arc, adjacent to the Sea of Japan on the San-in Coast of Honshu (Figure 1). The lavas erupted at Daisen are compositionally bimodal, consisting of marginally tholeiitic basalts and calcalkaline intermediate lavas, separated by a ∼10 wt% silica gap. The first lavas erupted at Daisen are magnesian olivine-phyric basalts [Tamura et al., 2000], appearing at 1.21 ± 0.16 Ma (K-Ar date [Uto, 1989]). These are followed at 0.96 ±0.06 Ma (K-Ar date [Tsukui et al., 1985]) by the Tsubanukiyama dacite dome, which marks the beginning of a long sequence of eruptive events producing voluminous tephras and tuff breccias, aphyric andesites, dacite domes, and porphyritic intermediate lava flows, with no further surface expression of basaltic magmatism. Volcanic activity continued until 17.4 (±1.0) Ka (14C date [Nomura et al., 1995]), culminating in the event that produced the Misen dacite dome and associated pyroclastic flows. The volcanic complex sits unconformably atop Jurassic to Cretaceous granites and gneisses [Ishiga et al., 1989; Tsukui, 1985]. The total amount of material erupted at the Daisen volcanic field is estimated at >120 km3 [Tsukui, 1984, 1985]. A list of samples used in this study with locations and published ages is given in Table S1 (supporting information).1
 Major element concentrations, trace element concentrations, and Sr, Nd, and Pb isotope ratios were determined at the Pheasant Memorial Laboratory (PML), Institute for Study of the Earth's Interior (ISEI), Okayama University at Misasa [Nakamura et al., 2003].
3.1. Major Elements
 Major and minor element compositions were determined by X-Ray Fluorescence Spectroscopy (XRF) using a Phillips PW2400 spectrometer. Glass beads were prepared by fusing 500 mg of powdered sample using 5 g of lithium tetraborate as a flux [Takei, 2002]. Loss on ignition (L.O.I.) was determined gravimetrically. All samples were prepared and analyzed in duplicate. In this study, the quality of the major element analyses assessed by totals is 100 ± 0.8%.
3.2. Trace Elements
 Trace element concentrations of whole rock samples were measured using an Agilent 7500cs ICP-MS, following the procedures of Makishima and Nakamura  and Lu et al. . For dacitic samples, a Teflon® bomb decomposition method (4 days at 245°C) was used in order to decompose acid-resistant minerals such as zircon [Makishima et al., 1999]. Magnesium was added in order to suppress AlF3 formation during the high-pressure acid digestion procedure [Takei et al., 2001]. All samples were digested and analyzed in duplicate to ensure reproducibility. The replicate analyses were within 3–5% relative difference. Uranium and thorium concentrations of selected samples (2051801, 3052002, 3052004, 3052802, 3052803, SAN1, SAN2, 2060707, 2060710, 2060711, 2060713, and 2080501) were determined by isotope dilution thermal ionization mass spectrometry (ID-TIMS) using a Finnigan MAT262 (“INU”) [Yokoyama et al., 1999, 2001, 2003]. Replicate analyses by this method were within 1% relative difference.
3.3. Sr, Nd, and Pb Isotope Ratios
 Isotopic analyses of Sr, Nd, and Pb were performed by thermal ionization mass spectrometry (TIMS) in static multicollection mode using either the Finnigan MAT262 (“TARO”) or Thermo TRITON at the Pheasant Memorial Laboratory. Procedures followed the methods of Yoshikawa and Nakamura  for Sr, Nakamura et al.  for Nd, and Kuritani and Nakamura  for Pb (double spike method). Typical analytical reproducibility was 0.005, 0.005, and 0.02%, for Sr, Nd, and Pb isotopes, respectively. Isotopic fractionation during analysis was corrected using 86Sr/88Sr = 0.1194 and 146Nd/144Nd = 0.7219 as normalization factors.
 During analysis using the Finnigan MAT262, the measured ratios of standard materials were 87Sr/86Sr = 0.710180 ± 42 (2σ, n=10) for NIST987 and 143Nd/144Nd = 0.511733 ± 18 (2σ, n=10) for the in-house standard PML-Nd, which is equivalent to 143Nd/144Nd = 0.511869 for the La Jolla standard reference material. During analysis using the Finnigan TRITON, the measured ratios of standard materials were 87Sr/86Sr = 0.710273 ± 19 (2σ, n=7) for NIST987 and 143Nd/144Nd = 0.511749 ± 9 (2σ, n=6) for PML-Nd, which is equivalent to 143Nd/144Nd = 0.511885 for La Jolla. Reported εNd values were calculated using 143Nd/144Nd CHUR = 0.512638 after normalization by La Jolla Nd = 0.511860. The NIST981 Pb standard yielded averages (n=5) of 206Pb/204Pb = 16.941 ± 0.002, 207Pb/204Pb = 15.499 ± 0.002, and 208Pb/204Pb = 36.726 ± 0.005 (all 2σ). In Table 1, all data are adjusted for instrumental discrimination to 87Sr/86Sr= 0.710240 for NIST987, 143Nd/144Nd = 0.511860 for La Jolla [Makishima and Masuda, 1994], and 206Pb/204Pb = 16.941, 207Pb/204Pb = 15.499, and 208Pb/204Pb = 36.723 for NIST981 [Kuritani and Nakamura, 2003].
Table 1. Major and Trace Element Concentrations and Isotope Ratios in Selected Daisen Lavas
 Powdered rock samples were mixed with 229Th and 233U spikes and dissolved using a mixture of HF and HClO4, following the method of Yokoyama et al. . A Finnigan MAT 262 (“INU”) was used to determine 230Th/232Th and 234U/238U. Detailed descriptions of the ion exchange procedures, filament loading, and analytical conditions are given in Yokoyama et al. [1999, 2001, 2003]. The precision of the 230Th/232Th ratio within a given run ranged from 0.3 to 0.8% (average 0.5%, 2σmean), and the reproducibility of Geological Survey of Japan (GSJ) standards [Imai et al., 1995] is 0.77% (2σ, n=6) for JB-3 (basalt) and 0.72% (2σ, n=7) for JR-2 (rhyolite). The accuracy of our Th isotope analysis was confirmed by repeated measurements of AThO (230Th/232Th = 5.440 ± 0.039 ×10−6, 2σ, n=5), and we obtained data for this standard consistent with the average of the previously published data (230Th/232Th = 5.475 ± 0.039 ×10−6, 2σ, n=7 [Williams et al., 1992; McDermott et al., 1993; Sigmarsson et al., 1998]). Replicate analyses of a standard rock sample JB-2 (tholeiitic basalt of Izu-Oshima, Japan) obtained from the Geological Survey of Japan gave (230Th/232Th) = 1.250 ± 0.013, Th = 0.2561 ± 0.0022 (µg/g), (234U/238U) = 1.001 ± 0.002, and U = 0.1536 ± 0.0013 (µg/g), respectively (errors are 2σ: Yokoyama and Nakamura, 2004; Yokoyama et al., 2006). The 234U/238U within-run precision ranged from 0.1 to 0.4% (2σmean), and the reproducibility of GSJ standards was 0.16% (2σ, n = 5) for JB-3 and 0.30% (2σ, n = 5) for JR-2. The 238U/232Th within-run precision ranged from 0.3% to 0.5% (2σmean). The accuracy of our U isotope analyses in 233U-spiked runs was confirmed by measuring the U standard solution (NBL-145) with spiked 233U (234U/238U = 5.284 ± 0.011 ×10−5, 2σ, n=10), demonstrating good agreement with previous work [Chen et al., 1986; Stirling et al., 1995; Luo et al., 1997; Cheng et al., 2000; Yokoyama et al., 2001].
4.1. Major Elements
 Two distinct lava types are present at the Daisen volcanic complex: a basaltic series with a very limited range of SiO2 from ∼49 to 50 wt.%, and an andesite-dacite series with SiO2 ranging from 58 to 66 wt.% (Table 1 and Figure 2, Figure S1, supporting information). Although the basalts plot (barely) within the subalkaline sector of the total alkalis versus silica diagram (Figure 2a, after LeBas et al. ), they have relatively high total alkali contents (∼4.5% K2O+Na2O at ∼50% SiO2) compared to typical arc basalts. The basalt series plots directly on top of the tholeiite-calcalkaline discrimination line on the plot of FeO*/MgO versus SiO2 (Figure 2b, after Miyashiro ), distinct from the clearly calcalkaline andesite-dacite series. Overall, the basalt series displays a limited range of major and minor element concentrations, including TiO2, MgO, NiO, and Cr2O3, while the andesite-dacite series shows insignificant variation for these elements. In contrast, the andesite-dacite series displays a significant range of concentrations of elements commonly affected by fractional crystallization, including Al2O3, CaO, Na2O, and K2O, while these elements are essentially constant in the basalts. These major element trends are consistent with the lack of plagioclase phenocrysts in the basalts [Tamura et al., 2000], compared to extensive plagioclase crystallization in the dacites [Tamura et al., 2003].
4.2. Trace Elements
 Trace element concentrations are presented in Table 1. Both the intermediate lavas (Figure 3a) and the basalts (Figure 3b) show evidence for enrichment of Pb, Sr, and Li, and depletion of Nb (and Ta), when plotted on a mid-ocean ridge basalt (MORB)-normalized “spider” diagram. However, these enrichments and depletions are much more pronounced in the intermediate series. In contrast to the basalts, the intermediate series lavas are more enriched in U, Th, Zr, and Hf, and distinctly depleted in heavy rare earth elements (HREE). Very small positive Eu-anomalies (<6% in basalts and <15% in andesites and dacites) are observed, and both series are depleted in HREE when plotted on a chondrite-normalized diagram, although the HREE depletion in the intermediate lavas is more pronounced (Figure 3c). Granitic country rocks are enriched in Rb, Th, U, Pb, and HREE, but depleted in Ba, Sr, and Eu, relative to the intermediate series (Figures 3a and 3c).
4.3. Sr, Nd, and Pb Isotope Ratios
 Strontium and neodymium isotope ratios are presented in Table 1, and 87Sr/86Sr is plotted against εNd in Figure 4a. The basalts have a relatively limited range of isotope ratios, with 87Sr/86Sr ∼0.7050–0.7051 and εNd = 0.5–1.2, while the intermediate series lavas have a moderate range of 87Sr/86Sr ∼0.7046–0.7052 and εNd = 1–3. The intermediate suite forms a more or less linear array with a negative slope on a plot of 87Sr/86Sr versus εNd. The basalts are offset to lower εNd relative to the intermediate lavas, and fall at the upper limit of 87Sr/86Sr values for the Daisen volcanics.
 Lead isotope ratios are presented in Table 1 and Figures 4b and 4c. The basalts and intermediate suite form subparallel arrays with positive slopes on plots of both 206Pb/204Pb versus 207Pb/204Pb and 206Pb/204Pb versus 208Pb/204Pb. The two arrays are clearly offset in 206Pb/204Pb versus 208Pb/204Pb, but only very subtly differentiated in 206Pb/204Pb versus 207Pb/204Pb. The basalts are offset to higher 208Pb/204Pb and lower 206Pb/204Pb relative to the andesites and dacites. All Daisen lavas fall well above the Northern Hemisphere Reference Line (NHRL; [Hart, 1984]).
 Uranium-thorium activity ratios were determined for the three youngest volcanic features at Daisen: Misen (17 Ka), Karasugasen (60 Ka), and Sankoho (20–90 Ka). The activity ratios (230Th/232Th) and (238U/232Th) are presented in Table 1 and shown in Figure 5. Initial thorium activity (230Th/232Th)i is calculated using eruption ages determined by K-Ar dating (Karasugasen: Kimura ; Sankoho: Tsukui et al. ) or 14C dating of charcoal (Misen: Nomura et al. ) and reported in Table 1. Because there is such a disparity of dates for Sankoho (20–90 Ka), (230Th/232Th)i is calculated for the minimum and maximum ages and reported as a range. The Misen and Sankoho lavas are characterized by small (up to 9%) initial 238U-excesses, while the Karasugasen lavas are characterized by very large (up to 60%) initial Th-excesses. The Karasugasen lavas display a relatively large range of (230Th/232Th)i, varying from 0.748 to 0.873. This variability may reflect a heterogeneous source or low-temperature alteration (although the anomalous sample has (234U/238U) = 1.0, which suggests that low-temperature disruption of the U-series is minimal). It is interesting to note that the Sankoho and Misen lavas plot around a single line on a plot of (238U/232Th) versus (230Th/232Th) (Figure 5), suggesting that these may originate from a single batch of magma.
5.1. Daisen Intermediate Lavas
 All of the Daisen intermediate lavas display enrichment of Pb and Sr, and depletion of Nb, on a MORB-normalized “spider” diagram (Figure 3a). These geochemical features are typical of arc magmas and indicate a contribution of material from the subducted slab (altered oceanic crust + subducted sediment). Compared to intermediate lavas erupted in the Northeast Japan arc [Kimura and Yoshida, 2006], the Daisen intermediate lavas have more subtle enrichments in fluid-mobile elements such as Ba and Pb (Figure 3a), which is reflected in their lower average Ba/Nb (65 at Daisen; 110 in NEJ) and higher average Ce/Pb (5 at Daisen; 3 in NEJ) ratios, suggesting a weaker dehydration flux and/or a stronger slab-derived melt contribution in the Southwest Japan arc. The most distinctive feature of the Daisen intermediate lavas is their depletion in middle and heavy REEs (Figures 3a and 3c), consistent with interaction with amphibole and/or garnet.
 In contrast to lavas from the Northeast Japan and Izu arcs [Moriguti et al., 2004], distinct MORB and mantle sources are not evident in the isotopic characteristics of the Daisen intermediate lavas (e.g., Figures 4e and 4f). Rather, the intermediate lava series is best described by two-component mixing of Philippine Sea Plate MORB [Hickey-Vargas, 1991, 1998] with Nankai Trough/Philippine Sea Plate sediment [Terakado et al., 1988; Ishikawa and Nakamura, 1994; Shimoda et al., 1998]. This may be because the Philippine spreading center is located so close to the Nankai Trough that the mantle source of the Philippine Sea Plate MORB is very similar to the mantle that comprises the wedge beneath Southwest Japan. Thus, although there may be separate processes supplying material from the subducted oceanic crust and the mantle wedge, these contributions could be isotopically indistinguishable. It is also conceivable that the intermediate magmas at Daisen are derived directly from melting of the subducted slab, with insignificant input from the mantle wedge.
5.2. Evidence for Slab Melting?
 The intermediate lavas of the Daisen volcanic field are rich in Al2O3 and poor in MgO, consistent with the slab melt characteristics given by Defant and Drummond , namely: ≥ 56% SiO2, ≥15% Al2O3, and <3% MgO (rarely >6%). The trace element characteristics of the Daisen intermediate lavas are also consistent with slab melting, with Sr/Y = 70–105 (Y <10 ppm) and La/Yb (normalized to chondrite) = 21–29 (Yb <1 ppm) (Figures 6a and 6b, respectively). These trace element ratios are interpreted to reflect melting in the garnet stability field, with the HREEs and Y being incorporated into residual garnet [Kay, 1978; Defant and Drummond, 1990]. Magmas with these characteristics are commonly referred to as adakites, in reference to their first documented occurrence on the island of Adak in the Aleutians [Kay, 1978]. According to the classification scheme of Martin et al. , the Daisen intermediate lavas would be categorized as high-silica adakites (HSA), defined by their relatively high SiO2 contents (>60%), low MgO (<4%), and Sr <1000 ppm. Such lavas have been interpreted as direct melts of subducted oceanic crust that have undergone only very limited reaction with mantle peridotite before being erupted at the surface.
 A slab melting origin has previously been proposed for Daisen and other Quaternary Southwest Japan Arc volcanic centers [Morris, 1995]. The Southwest Japan Arc is an ideal location for slab melting due to the subduction of the very young (∼20 Ma), and hence relatively hot, Philippine Sea Plate. This results in a slab surface temperature of ∼800°C at subarc depths beneath the Southwest Japan Arc, in contrast to a slab surface temperature of ∼500°C beneath the Northeast Japan Arc [Peacock and Wang, 1999]. The model slab surface temperature for Southwest Japan is well above the wet basalt solidus of Schmidt and Poli , suggesting that partial melting of the slab is likely.
 The Misen and Sankoho lavas from Daisen have small initial 238U-excesses, while the Karasugasen lavas have very large initial 230Th-excesses (Figures 6c and 6d). This in in contrast to the moderate-to-large 238U-excesses typically observed in arc lavas [e.g., Gill and Williams, 1990; McDermott and Hawkesworth, 1991; Hawkesworth et al., 1997; Turner et al., 1997], including those from the Izu Arc [Yokoyama et al., 2003], but consistent with 230Th-excesses observed in some arc lavas that display adakitic trace element signatures [e.g., Sigmarsson et al., 1998]. Thorium excesses or diminished U excesses are consistent with melting in the presence of garnet due to retention of U preferentially over Th, as DU>DTh in garnet with respect to silicate melt [Beattie, 1993; LaTourrette et al., 1993; Klemme et al., 2002]. Tokunaga et al.  propose a relationship between Sr/Y and (238U/230Th) for Daisen dacites, observing that Karasugasen is characterized by higher Sr/Y and (238U/230Th) < 1, while Misen has somewhat lower Sr/Y and (238U/230Th) > 1. This relationship is less clear in our data set, but still consistent with our observations (Figure 6c). Tokunaga et al.  attribute the U-excess in Misen lavas to interaction of slab melt with fluid-metasomatized mantle. Although some exchange with the mantle is probably inevitable, the very low MgO content (<3%) in the Misen lavas argues against extensive mantle interaction. We propose instead that the variations in (238U/230Th) and Sr/Y may reflect varying residual assemblages during slab melting and perhaps during magma storage and evolution in the lower crust. Note that all of the Daisen lavas have low (238U/230Th) compared to lavas from the Izu Arc [Yokoyama et al., 2003], suggesting that U has been retained at depth relative to a “normal”, fluid-rich arc (Figure 6d).
5.3. Lower Crustal Interaction
 An alternative process for generating adakitic trace element signatures in arc magmas without calling upon slab melting is by equilibration with garnet in thick lower crust [Kay and Kay, 1991; Macpherson et al., 2006; Davidson et al., 2007]. Fractional crystallization of pyrope-rich garnet, in the absence of plagioclase, can lower the Mg-content of arc magmas, while increasing Sr/Y and REE fractionation. The crust beneath Daisen is ∼31 km thick [Salah and Zhao, 2004], slightly less than the minimum depth required for garnet stability in a mafic bulk composition. However, Zellmer et al.  argue that the relatively dry nature of the Daisen basaltic magmas (∼1–2 wt.% H2O, compared to ∼5 wt.% in Kyushu) could expand the garnet stability field such that fractional crystallization of garnet would be possible in the crust beneath Daisen, assuming that the basalts are the parental magmas to the intermediate series. We will address the validity of this assumption in Section 'Daisen Basalts: Parental to the Intermediate Lavas?'. In general, magma injection and storage at the base of the crust will be characterized by fractional crystallization accompanied by assimilation of or hybridization by partial melts of the lower crust [DePaolo, 1981; Hildreth and Moorbath, 1988; Annen et al., 2006]. In order to assess the extent of lower crustal interaction at Daisen, we compare the isotopic compositions of the intermediate lavas with those of lower crustal xenoliths from Oki-Dogo in Southwest Japan (Figures 4d–4f).
 Oki-Dogo is a volcanic island in the Sea of Japan, ∼90 km north of Daisen. Moriyama  studied the isotope geochemistry of basalt-hosted xenoliths from Oki-Dogo. Lower crustal mafic and ultramafic xenoliths from Oki-Dogo can be distinguished from cumulate xenoliths on the basis of their granoblastic textures and distinct isotopic compositions. The isotopic characteristics of the lower crustal xenoliths are distinct from those of any other component contributing to magma-genesis at Daisen, such that lower crustal assimilation should be evident in the isotopic trends. However, the isotopic trends, most notably the trend in 206Pb/204Pb versus 208Pb/204Pb (Figure 4f), are inconsistent with assimilation of Oki-Dogo lower crust. Assuming that the lower crust beneath Daisen is comparable to that beneath Oki-Dogo, there does not appear to be significant interaction with lower crust by the Daisen magmas. An alternative possibility for lower crustal composition is discussed in Section 'Implications for Sediment Subduction'.
5.4. Subducted Sediment Contribution
 Isotopically, the Daisen intermediate lavas differ from typical slab melts as defined by Kay  and Defant and Drummond  by virtue of their relatively radiogenic Sr isotope ratios. The Daisen lavas have 87Sr/86Sr = 0.7046–0.7052, whereas typical adakites are defined by 87Sr/86Sr < 0.7040. The single granitic country rock we analyzed does not make an appropriate end-member for the Sr-Nd isotopic trends observed in the intermediate lavas, as was also observed by Tamura et al. , and a limited role for country rock assimilation is particularly evident in the 208Pb/204Pb versus 206Pb/204Pb trend (Figure 4f). Instead, the more radiogenic Sr at Daisen relative to other adakitic centers can be attributed to a relatively large mass contribution from subducted sediment.
 Further evidence for a significant sediment contribution comes from Th isotope systematics. In the Daisen lavas, current (230Th/232Th) values range from ∼0.72–0.84 (Figure 5a). The Misen and Sankoho trends intersect the equiline in close proximity to the (230Th/232Th) value of 0.76 calculated for global subducted sediment (GLOSS; [Plank and Langmuir, 1998]), assuming U/Th = 0.24 and a starting condition of secular equilibrium. The low (230Th/232Th)i in the Daisen lavas correlates with high Th/La (an index of sediment contribution [Plank, 2005]), in contrast to higher (230Th/232Th) and lower Th/La in young lavas from the Izu Arc (Figure 5b). Note that the sediment component at Daisen is more pronounced than in even the most strongly sediment-influenced Mariana arc lavas [Elliott et al., 1997]. These observations are consistent with a significant input of U and Th from subducted sediment beneath Daisen.
5.5. Melting and Mixing Models
 We have calculated trace element abundances in a theoretical magma generated by mixing partially melted (F=15%) Philippine Sea Plate MORB with partially melted (F=30%) subducted sediment, in order to determine the relative inputs of slab melt and sediment melt to the Daisen intermediate lavas. Partition coefficients for eclogite and sediment were taken from Klemme et al.  and Johnson and Plank , respectively. Garnet amphibolite partition coefficients were calculated using the amphibole-dacite partition coefficients of Sisson  for REE, and Ewart and Griffin  for all other elements, combined with the garnet-melt partition coefficients of Klemme et al.  at a ratio of 40:60 amphibole:garnet. The oceanic crust composition used was that of Philippine Sea Plate MORB (averaged Shikoku, Parece Vela, and West Philippine Basin glasses [Hickey-Vargas, 1998]), and the sediment bulk composition was that of Philippine Plate pelagic sediment [Shimoda et al., 2003]. The incompatible element concentrations of the Daisen intermediate lavas are generally well fit by a mixture of ∼25% partially melted sediment with ∼75% partially melted oceanic crust (Figure 7a). However, comparison of the eclogite melting (subducted slab) and amphibolite melting (lower crust) models is not straightforward. The high Sr concentrations that are the hallmark of the Daisen intermediate magmas are only reproduced by partial melting of MORB at eclogite-facies conditions. Conversely, the REE patterns of the Daisen magmas are best fit by melting in the presence of amphibole. Likewise, the Zr/Sm versus Nb/Ta diagram of Foley et al. , designed to discriminate between eclogite and amphibolite melting, is inconclusive—the Daisen lavas do not feature the very high Zr/Sm and low Nb/Ta characteristic of low-degree amphibolite melts, yet they display a range of Nb/Ta greater than that predicted for rutile-free eclogite melts (Figure 7b). Thus a more complex, multistage process (or more appropriate partition coefficients) may be necessary to adequately model the trace element characteristics of the Daisen intermediate lavas. A likely scenario is that the magmas were initially formed by melting of the slab in the eclogite stability field, and later reworked in the lower crust, either by equilibration with amphibolite or by fractional crystallization of amphibole. We will proceed on the assumption that the mixing ratio of ∼3:1 partially melted oceanic crust to partially melted sediment best describes the overall abundances of incompatible elements in the magmas, regardless of residual assemblage.
 Parabolic mixing curves for 87Sr/86Sr versus εNd are poorly constrained due to the lack of a single melting model to simultaneously reproduce both the Sr and Nd concentrations in the magmas. Nonetheless, mixing of partial melts of Philippine Sea Plate MORB and sediments using imposed concentration ratios can reproduce the trend observed in the Daisen lavas. Isotope ratios of endmembers are taken from Philippine Sea Plate MORB (Shikoku, Parece Vela, and West Philippine Basin glasses [Hickey-Vargas, 1991]) and the Shimanto Shale, an exhumed analog for subducted Nankai Trough sediments [Terakado et al., 1988]. Assumed concentration ratios for the crustal melt relative to the sediment melt are 2:1 for Sr and 1:2 for Nd. The resulting curve is consistent with mixing ∼70% oceanic crust partial melt with ∼30% subducted sediment partial melt (Figure 4d). Lead isotope mixing models, which are insensitive to the interelement fractionation issues that complicate the Sr-Nd model, also indicate that ∼20–30% partially melted Nankai subducted sediment (Shimanto Shale [Ishikawa and Nakamura, 1994]) mixed with partially melted Philippine Sea Plate MORB [Hickey-Vargas, 1991] can describe the isotopic variations observed in the intermediate lavas for either eclogite- or amphibolite-facies partial melting (Figures 4e and 4f).
5.6. Implications for Sediment Subduction
 The significant amount of subducted sediment (∼20–30%) in the Southwest Japan arc source, as suggested by radiogenic isotope mixing models (Figures 4d–4f), and reflected in the low (230Th/232Th) ratios in the Daisen lavas (Figures 5a and 5b), is interesting given that the modern Nankai Trough is largely an accretionary margin. At present, a relatively modest 350 m of pelagic clay located beneath the decollement in the Nankai accretionary prism is believed to make its way into the subarc region [Plank and Langmuir, 1998]. Why then is the apparent sediment contribution at Daisen and other volcanic centers in Southwest Japan so strong?
 We offer a few speculative ideas for processes by which a significant quantity of sediment might have accumulated in the mantle or at the base of the crust beneath the Southwest Japan arc since the initiation of Shikoku Basin subduction in the early Miocene [Tatsumi, 1983]. Recent modeling suggests that subducted sediments of thickness >100 m will detach from the slab to form buoyant diapirs at subarc depths, on a timescale of <1 Myr [Behn et al., 2011]. For the Southwest Japan arc (Nankai), detachment is predicted to initiate at temperatures between 650 and 675°C [Behn et al., 2011], which corresponds to slab surface temperatures ∼75 km trenchward of the Quaternary arc [Peacock and Wang, 1999]. As these metasedimentary diapirs rise through the mantle wedge, they experience partial melting, contributing a sediment melt component to the mantle wedge. The restites of melt extraction may then be relaminated at the base of the arc crust to form a felsic lower crustal reservoir [Hacker et al., 2011]. The aluminous composition of the metasediments would expand the garnet stability field relative to more mafic lower crustal compositions, putting the base of the Southwest Japan arc crust, at ∼31 km depth [Salah and Zhao, 2004], within the garnet stability field. Thus a felsic layer at the base of the crust could contribute to Quaternary magma-genesis in the Southwest Japan arc, which would be consistent with the distinctive trace element and isotopic compositions observed at Daisen.
 Alternatively, seismic studies have found evidence for ridge subduction and forearc erosion at the base of the accretionary prism at the Nankai Trough [Bangs et al., 2006]. A 1 km high sedimentary packet in the wake of a subducted ridge could detach from the slab to form a sediment plume or diapir, accumulating to form a sediment-rich reservoir in the upper back-arc mantle [Currie et al., 2007]. If such processes were active in the Miocene, when the active arc was ∼100 km trenchward of the modern Southwest Japan arc, we would expect the remains of the shallow mantle sediment reservoir to underlie modern-day Daisen and the rest of the Quaternary arc. Partial melts of the Philippine Sea Plate could then stall in a felsic lower crustal hot zone evolve toward lower MgO and higher Al2O3, with enriched sediment-like radiogenic isotope signatures, without erasing (or conceivably while enhancing) the garnet signature in the trace element characteristics.
5.7. Daisen Basalts: Parental to the Intermediate Lavas?
 The basalts erupted at the Daisen volcanic field are generally comparable to the most enriched basalts erupted in the back-arc region of the Northeast Japan arc [Shibata and Nakamura, 1997], with the most notable differences being lower Cs, Rb, U, and HREEs, and slightly higher Nb (Figure 3b). The lower large ion lithophile to high field strength element ratio (LIL/HFSE) is characteristic of Quaternary and Tertiary alkali basalts from Southwest Japan [Nakamura et al., 1990; Uto and Tatsumi, 1996]. Relative to the intermediate lavas, the basalts at Daisen are skewed isotopically toward an EM1-like mantle endmember also observed regionally in alkali basalts (Figures 4d–4f [Nakamura et al., 1990; Tatsumoto and Nakamura, 1991]). It has been argued that this enriched mantle component reflects melting of ancient subduction-enriched lithosphere [Tatsumoto and Nakamura, 1991; Uto et al., 1994; Ikeda et al., 2001] and/or involvement of a mantle plume in the opening of the Sea of Japan [Nakamura et al., 1985, 1989, 1990]. Although the Daisen basalts are marginally subalkaline, the relatively high total alkalis in these lavas (Figure 2a), together with their enriched incompatible element compositions (Figure 3b), are consistent with a small contribution from an enriched mantle source. Given the location of Daisen on the San-in Coast of Southwest Honshu, it is reasonable to expect that the mantle source for these basalts could contain a Japan Sea enriched mantle component.
 The relationship between the basalts erupted around Daisen and the intermediate lavas that comprise the Daisen volcanic centre has been debated. The basalt flows located on the southwest flank of Daisen have been grouped with other monogenetic basaltic centers of similar age further to the southwest (e.g., the “Yokota” group of Iwamori ), which are clearly not directly related to the intermediate volcanism at Daisen. On the other hand, Tamura et al.  argued that the two magma types are closely related, and the basalts should be considered part of the Daisen volcanic complex based on the overall isotopic (Sr-Nd) similarity to the intermediate lavas. A close look at the isotope ratios reveals that the basalts have 87Sr/86Sr values equal to the most radiogenic values observed in the Daisen intermediate lavas, and εNd lower than any observed in the intermediate suite (Figures 4a and 4d). This observation is difficult to reconcile with derivation of the intermediate magmas from the observed basalt composition by a typical process of assimilation and fractional crystallization, which would tend to increase radiogenic Sr in the andesites and dacites relative to the basalts, assuming the crustal assimilant has a radiogenic 87Sr/86Sr composition. Also, the basalts are offset from the linear trend of the intermediate lavas toward lower εNd (Figure 4a and 4d). The difference is even more apparent when looking at 206Pb/204Pb versus 208Pb/204Pb (Figures 4c and 4f), as the basalts appear to lie along a trendline parallel to that observed in the andesite-dacite suite. In both cases, the basalts are shifted toward the EM1-like enriched mantle endmember observed in many alkaline basalts and mantle xenoliths from around the Sea of Japan. Thus, although the basalts and intermediate lavas erupted at the Daisen volcanic center may share some of the same components (e.g., subducted Nankai Trough sediment), there is no clear evolutionary path to suggest that the basalts are parental to the andesites and dacites.
 Major element, trace element, and isotope ratio evidence for basalts and intermediate lavas erupted at the Daisen volcanic field over the past ∼1.2 Myr suggests that: 1) all of the lavas erupted at Daisen contain a significant contribution from subducted sediment; 2) the intermediate lavas have equilibrated with garnet at some point during magma genesis and/or magmatic evolution; and 3) the basalts erupted at 1.2 Ma contain a contribution from an EM1-like mantle source and are not parental to the intermediate lavas that have erupted over the past ∼500 Kyr.
 Isotopic similarities between Nankai Trough sediments and Miocene volcanics in the forearc of the modern Southwest Japan arc suggest that significant sediment was added to the mantle at the time of initiation of Shikoku Basin subduction at ∼20 Ma [Ishikawa and Nakamura, 1994]. We speculate that remnants of this past sediment subduction may have been relaminated at the base of the arc crust and/or contributed to a broadly sediment-contaminated mantle wedge. The simplest explanation for the genesis of the intermediate lavas is that they originated by direct melting of the hot, young subducted slab (both altered Philippine Plate oceanic crust and subducted Nankai Trough sediment). Slab melting may have been enhanced by interaction of hot upwelling mantle, possibly a plume, with the leading edge of the slab. Mantle upwelling would be consistent with the widespread appearance of EM1-like alkali basalts in the region. The intermediate magmas may have experienced assimilation and fractional crystallization in a lower crustal hot zone dominated by garnet-bearing metasedimentary rocks, and apparently had some interaction with amphibole prior to eruption. The spacing in time of 20 Kyr between the basalt eruption and the first appearance of intermediate lavas is consistent with the amount of time predicted for generation and mobilization of intermediate lavas in a lower crustal hot zone following addition of hot, mantle-derived basalt [Annen et al., 2006]. Sediment assimilation at the base of the arc crust could explain the high Al2O3, low MgO, HREE depletion, and sediment-like radiogenic isotope compositions of the intermediate lavas erupted at the surface. A schematic diagram representing our conceptual model for magmagenesis at Daisen is shown in Figure 8.
 The basalts erupted around Daisen have isotopic characteristics that are broadly similar to the Daisen intermediate lavas, but are distinctly offset toward an EM1-like mantle component similar to that observed in widespread alkali basalts from around the Sea of Japan. The regional alkali basalts have previously been attributed to a mantle plume rising beneath the Sea of Japan [Nakamura et al., 1985, 1989, 1990]. However, the EM1 component at Daisen is strongly attenuated, suggesting that if the Daisen basalts are plume-derived, the plume component has been significantly diluted by entrainment of sediment-infused subarc mantle beneath Southwest Japan.
 In summary, the Daisen intermediate lavas are most likely generated by partial melting of the subducted Philippine Sea plate mixed with Nankai Trough sediment, while the basalts are generated by partial melting of hybrid mantle composed primarily of subarc mantle that has been metasomatized by sediment-derived melts, mixed with an ancient enriched mantle (EM1) component. Sediment relamination at the base of the arc crust could contribute to the strong sedimentary influence in the trace element and isotopic characteristics of the intermediate magmas.
 The authors would like to thank Akio Makishima, Chie Sakaguchi, and Takeshi Kuritani for assistance with trace element, Sr-Nd isotope, and Pb isotope analyses, respectively. The samples analyzed in this study were collected by Sakiko Terui. Paul Morris is acknowledged for providing two additional samples from Sankoho. The authors thank Brian Jicha and Weidong Sun for their constructive reviews. This work was supported by a “Center of Excellence in the 21st Century in Japan” (COE-21) postdoctoral fellowship to MDF.