Possible partial melting and production of felsic melt in a Jurassic oceanic plateau of the Izanagi Plate: Insights from 159 Ma plagiogranites from northern Japan

ABSTRACT A small tonalite – dacite body has been discovered in the Nakanogawa Group, Hidaka Belt, northern Japan, which is a Palaeogene subduction complex formed in the palaeo-Japan trench along the northeastern margin of Eurasia. The tonalites are characterized by extremely low K2O (0.2–0.3 wt.%) and relatively flat chondrite-normalized rare earth element patterns (La/Yb[N] ~2.3), similar to plagiogranites in ophiolites. The major-and trace-element characteristics are consistent with low-pressure partial melting of oceanic crust outside the garnet stability field. Zircon U – Pb dating of the tonalite yielded a weighted-mean 206Pb/238U age of 159.1 ± 1.6 Ma. Late Jurassic oceanic plateau-type basaltic rocks with a small amount of plagiogranitic rocks are distributed sporadically in subduction complexes along the palaeo-Japan arc – trench system. A similar zircon U – Pb age (151.6 ± 1.8 Ma) has been reported for greenstone in the Cretaceous subduction complex along the palaeo-Kuril arc – trench system. The tonalite – dacite block was probably derived from the subduction complex at the junction of the palaeo-Japan and -Kuril arc – trench systems. Thus, the greenstones and the tonalite body were likely part of a large oceanic plateau, which formed at the Izanagi – Pacific–Farallon ridge triple junction during the Late Jurassic – Early Cretaceous.


Introduction
Subduction complexes comprise exotic blocks accreted by tectonic transfer of material from the subducting slab to the overlying plate.This material typically includes oceanic crust, fragments of seamounts or oceanic plateaus, pelagic sedimentary rocks such as red shale, bedded chert, and limestone, and terrigenous sedimentary rocks such as sandstone and mudstone (e.g.Isozaki et al. 1990;Kusky et al. 2013;Safonova and Santosh 2014;Wakita 2015).Greenstones are characteristic metamorphosed mafic igneous rocks in subduction complexes.Petrological and geochemical investigations of these greenstones can provide important insights into oceanic magmatism within the consumed plate, the origin of the accretionary complex, and growth of the continental margin (Ueda and Miyashita 2005;Ichiyama et al. 2008;Van der Meer et al. 2012;Sigloch and Mihalynuk 2013;Safonova and Santosh 2014;Safonova et al. 2015).In some cases, subduction complexes may contain felsic rocks such as dacite and granitic rocks, the study of which may lead to important insights into the original oceanic arc from which they are derived (e.g.Karig 1971, Ueda et al. 2005;Sigloch and Mihalynuk 2013;Yamasaki and Nanayama 2018).
The Panthalassa Ocean was an oceanic domain surrounding the supercontinent Pangaea during the late Paleozoic and early Mesozoic (e.g.Van der Meer et al. 2012;Boschman and van Hinsbergen 2016).The Izanagi Plate is assumed to have underlain the western Panthalassa Ocean to the east of Eurasia and to have been subducted beneath Eurasia during 60-50 Ma (e.g.Van der Meer et al. 2012, Matthews et al. 2016;Müller et al. 2016).The Izanagi Plate was eventually destroyed by subduction, but the subduction complexes of the circum-Panthalassa continental margins provide evidence for subduction-related volcanism within the Panthalassa Ocean (Isozaki et al. 1990;Van der Meer et al. 2012;Yamasaki and Nanayama 2018, and references therein).Ueda and Miyashita (2005) reported a narrow (<2 km) serpentinite-bearing belt of accretionary units containing remnant arc fragments in the Oku -Niikappu Complex of the Idonnappu zone of the Sorachi -Yezo Belt in northern Hokkaido, Japan (Figure 1a,b).The Oku -Niikappu Arc is overlain by the earliest Cretaceous (145-130 Ma) radiolarites that mark the minimum age of arc extinction and by mid-Cretaceous (~105-95 Ma) foreland basin clastic rocks that mark the timing of accretion (Ueda and Miyashita 2005).
Greenstones in subduction complexes of the circum-Panthalassa continental margins also provide evidence for anomalously large-scale, plateau-type volcanism within the Panthalassa Ocean during the Cretaceous (Tatsumi et al. 1998;Ichiyama et al. 2014).Large igneous provinces (LIPs) are the products of voluminous magmatism caused by mantle processes that are distinct from those forming oceanic crust at spreading ridges (e.g.Coffin and Eldholm 1994).Oceanic plateaus are thought to be products of large-scale mantle upwelling (e.g.Duncan and Richards 1991;Bryan and Ernst 2008), sometimes referred to as 'superplumes' (Larson 1991).It has been proposed that the South Pacific superplume in the Palaeo-Pacific Ocean was active at 150-90 Ma and was located in the equatorial to South Pacific region of the Izanagi Plate (e.g.Larson 1991;Tatsumi et al. 1998).Tatsumi et al. (1998) proposed that Late Jurassic superplume activity preceded the onset of the Cretaceous Normal Superchron , and that the superplume acted as a trigger for this global event.
The Mikabu Belt (also referred to as the Mikabu Greenstone or Mikabu Unit of the Sanbagawa terrane) in southwest Japan and the Sorachi -Yezo Belt in northern Hokkaido, Japan (Figure 1a,b), contain Late Jurassic oceanic basalts that are presumed to have been formed on an oceanic plateau (e.g.Sakakibara et al. 1999;Ichiyama et al. 2014;Safonova et al. 2016;Ueda 2016;Sawada et al. 2019).These rocks include those with distinctly high Nb/Y ratios at a given Nb/Zr ratio (Tatsumi et al. 1998) and picritic rocks formed at high (up to 1650°C) mantle potential temperatures (Ichiyama et al. 2014).The Sorachi -Yezo Belt extends to the Susunai and Aniva complexes on Sakhalin Island in the Russian Far East (Kimura et al. 1994;Tatsumi et al. 1998;Safonova and Santosh 2014 and references therein).Kimura et al. (1994) showed that accreted fragments of Late Jurassic oceanic plateau basalts in Japan can be traced back to the central part of Panthalassa using palaeomagnetic data and plate reconstructions.Based on the similar age of the plateau (i.e. the Sorachi Plateau) and Shatsky Rise, and the palaeogeography, Kimura et al. (1994) proposed that the Sorachi Plateau in the Sorachi -Yezo Belt is a missing twin of the Shatsky Rise in the modern Pacific Ocean.Furthermore, Kimura (1997) noted that the Mikabu and Sorachi -Yezo belts were a single large Jurassic oceanic plateau that accreted onto the Eurasian continental margin, based on the similar ages of both belts.Recently, Tominaga and Hara (2021) questioned the missing twin hypothesis based on geochemical and geochronological data for the Mikabu Unit and the Middle Jurassic to Early Cretaceous Chichibu accretionary complex.Tominaga and Hara (2021) suggested that the Mikabu basalts formed on older oceanic crust of the Izanagi Plate, which was located several thousand kilometres from the Pacific -Izanagi-Farallon triple junction, because the accretion age for the Mikabu Plateau was too young (ca.90 Ma) relative to estimates from the geological record (>110 Ma).
Previous research on superplume activity focused on the petrological and geochemical characteristics of primitive rocks, particularly picritic rocks, as these are direct products of the voluminous magmatism caused by large, hot plumes (Kimura et al. 1994;Tatsumi et al. 1998;Ichiyama et al. 2014); however, the felsic rocks accompanying such primitive rocks can provide significant insights into petrological processes.Compared with stepwise 40 Ar/ 39 Ar dating of basaltic rocks, which can be influenced by alteration (e.g.Walker and McDougall 1982;Jiang et al. 2021), zircon U -Pb dating of silicic rocks is simpler and provides more robust ages.Given these silicic lithologies, including tonalites and trondhjemites, are thought to be products of the direct melting of oceanic lithosphere, their existence is restricted to high heat flux regions, typically mid-ocean ridges.Similar conditions would be expected above the plume head region of a superplume, although such phenomena have been poorly documented.
We present petrological and geochemical data, including Sr -Nd-Pb isotopic compositions and zircon U -Pb ages, from a Late Jurassic plagiogranite in the Nakanogawa Group, Hidaka Belt.We discuss the origin of the plagiogranite body and petrogenesis of the felsic magma, and reconstruct a large Jurassic oceanic plateau and the magmatic processes associated with the Late Jurassic superplume.

Geological setting
Hokkaido, the northernmost major island of Japan, is located at the junction of two active island-arc -trench systems, the Northeast Honshu Arc -Japan Trench along the eastern margin of Eurasia, and the Kuril arc -trench system that extends from the Kamchatka Peninsula to eastern Hokkaido (Figure 1a).Hokkaido is divided into three major geotectonic units (western, central, and eastern Hokkaido) based on pre-Eocene basement geology.Western Hokkaido is the northern extension of Northeast Honshu, and consists of Jurassic subduction complexes with Cretaceous extrusive and intrusive rocks (e.g.Ueda 2016).Central Hokkaido consists of the Sorachi -Yezo Belt in the west and the Hidaka Belt in the east.The Sorachi -Yezo Belt is composed of Cretaceous clastic forearc basin sediments that overlie an ophiolite, and a Late Jurassic -Early Cretaceous siliceous sedimentary sequence, together with the Cretaceous high-pressure, high-temperature Kamuikotan metamorphic rocks (Ueda 2016).The Hidaka Belt consists of an early Palaeogene subduction complex that is dominated by clastic rocks, referred to as the Hidaka Supergroup (Nanayama et al. 1993(Nanayama et al. , 2019;;Ueda 2016).Eastern Hokkaido comprises a Late Cretaceous subduction complex associated with seamount fragments, called the Nikoro Group in the Tokoro Belt (e.g.Sakakibara et al. 1986), and Late Cretaceous -early Eocene forearc basin sediments and volcanic arc rocks, known as the Saroma and Nemuro groups of the Tokoro and Nemuro belts, respectively (Kiminami and Kontani 1983).

Sea of Japan
The Sorachi -Yezo Belt is divided into three units: the Kamuikotan, Sorachi, and Yezo units (e.g.Kimura et al. 1994 and references therein).The Sorachi and Kamuikotan units include rocks with a wide range of compositions, from tholeiitic basalt to picritic or strongly alkaline basalt.Tholeiitic basalts are dominant in the Sorachi unit, whereas picritic and alkaline basalts dominate the underlying Kamuikotan unit (Kimura et al. 1994;Ueda 2016, and references therein).Tatsumi et al. (1998) reported that basaltic rocks from the Kamuikotan unit have low Nb/Y and Nb/Zr ratios, similar to MORBs with Hawaiian-type geochemical signatures.However, Sakakibara et al. (1999) noted that the basaltic rocks in the Kamuikotan unit include alkaline basalts with HIMUlike trace-element signatures, which have a compositional spectrum through to MORB-like depleted basalts.Given that both MORB-and HIMU-like basalts are associated with Upper Jurassic to Lower Cretaceous chert and limestone without any terrigenous sediments, Sakakibara et al. (1999) suggested that the basaltic magmatism was formed by the South Pacific Superplume, and thus the Kamuikotan unit represents accreted fragments of the Late Jurassic Sorachi Plateau on the Izanagi Plate.
In southern -central Hokkaido, the Hidaka Supergroup is distributed along the Hidaka Mountains, and has been referred to as the Nakanogawa Group.Zircon U -Pb dating of sedimentary rocks from the Nakanogawa Group yields ages of 64.4 ± 1.0 to 48.8 ± 0.4 Ma (Nanayama et al. 2019).Nanayama (1992) noted that the rocks along the eastern margin of the Nakanogawa Group were highly deformed compared with other parts of the group.These deformed rocks, known as the Hiroo Complex, contain several allochthonous blocks of greenstone within a mélange (Nanayama 1992).The Daimaruyama assemblage (Figure 1c) is the largest (800 × 2000 m) greenstone block in the Hidaka Belt (Nanayama 1992).Yamasaki and Nanayama (2018) proposed that the Daimaruyama greenstones were submarine volcanic rocks that formed as a result of the intraoceanic subduction within the Izanagi Plate after the Early Cretaceous.They were eventually accreted onto the palaeo-Kuril arc -trench system at 57-48 Ma to form an allochthonous block in the Hiroo Complex, originally located on the landward side of the trench (Yamasaki and Nanayama 2018).In addition to the Daimaruyama greenstones, small greenstone blocks comprising ocean island basalt (OIB)-type alkaline volcanic rocks (e.g.Tachiiwa; Figure 1c) are distributed along the Hidaka Belt (Yamasaki and Nanayama 2020, and references therein).

Whole-rock major-and trace-element geochemistry
Whole-rock major-element compositions were measured using an X-ray fluorescence (XRF) spectrometer (Panalytical Axios) at the Geological Survey of Japan Laboratory (GSJ-Lab) in the GSJ, Tsukuba, Japan.The sample surfaces were scraped with a diamond disk to remove contamination from the rock saw.The samples were then cleaned with deionized water in an ultrasonic bath for >30 min and dried in an oven for >24 h.The dried samples were coarsely crushed in a tungsten carbide mortar and then ground in an agate mill.About 1.0 g of each powdered sample was weighed in a ceramic crucible and ignited in a muffle furnace for 2 h at 900°C.The difference in weight measured before and after heating was defined as the loss on ignition (LOI).Highdilution ratio (sample:flux = 1:10) glass beads were used for the XRF analysis.Whole-rock trace-element compositions were measured using a laser ablation -inductively coupled plasma -mass spectrometer (LA -ICP -MS), consisting of an Elemental Scientific Lasers NWR213 laser ablation system coupled to an Agilent 7700× quadrupole ICP-MS at the GSJ-Lab.The glass beads for the XRF analyses were used for the LA -ICP-MS analysis.XRF and LA -ICP-MS analytical methods followed those described by Yamasaki (2014) and Yamasaki and coarse-grained sandstone, respectively.Gm −01 and Gm −02 are greenish mudstone layers (Nanayama 1992).Greenstone complex is intruded into red bedded chert with Early Cretaceous Aptian-Albian (126-100 Ma) radiolaria (Nanayama 1992).Depositional age for the Hiroo Complex is after Nanayama et al. (2018Nanayama et al. ( , 2019)).
Yamashita (2016), respectively.The quality of the XRF and LA -ICP-MS analyses were monitored using analyses of U.S. Geological Survey (USGS) BCR−2 (Wilson 2000) and GSJ JA−1 (Imai et al. 1995) geochemical reference materials, respectively.The analytical results for the reference materials are listed in Table S1.

Amphibole and zircon major-and traceelement geochemistry
Major-element compositions of amphiboles were obtained using a J JEOL JXA-iHP200F electron microprobe analyser at the GSJ-Lab.A 12 nA beam current and 15 kV accelerating voltage were used, and the ZAF corrections were applied to all analyses.
Trace-element compositions of amphiboles and separated zircon grains prepared for U -Pb dating were measured using an LA -ICP-MS at the GSJ-Lab.Analytical methods followed those described by Yamasaki et al. (2015).For quality control, the analyses were calibrated to the NIST 615 glass analytical standard (results listed in Tables S2 and S3).

Whole-rock Sr, Nd, and Pb isotopic compositions
Approximately 100 mg of powdered rock sample was weighed, placed in a PFA vial, and dissolved in a mixture of 2.0 ml of ultra-pure hydrofluoric and nitric acids.The sample vial was immersed in an ultrasonic bath for 90 min, then placed in an electric oven at 90°C for 12 h, followed by heating to 120°C for 24 h for complete decomposition.The dissolved sample was evaporated until dry, and the residue was dissolved in 1.0 ml of 9.5 M ultra-pure hydrochloric acid and passed through an anion exchange resin bed polypropylene column (Bio-Spin® column; Bio-Rad Laboratories, U.S.A) packed with 1.0 ml of AG MP−1 M resin (200-400 mesh from Bio-Rad Laboratories, U.S.A) to remove Fe.The collected sample was dried, then dissolved in 1.0 ml of 3 M HNO 3 .Sr and Pb were purified using an extraction resin bed polypropylene column packed with 0.4 ml of Sr resin (50-100 μm, Eichrom Technologies, Inc., U.S.A) using the method described by Makishima et al. (2007Makishima et al. ( , 2008) ) and Wakaki et al. (2018).Sr and Pb were collected in 1.2 ml of 0.05 M HNO 3 and 5.0 ml of 6 M HCl, respectively.Nd separations were conducted using two-step column separation.First, light rare earth elements were separated from major elements using an extraction resin bed polypropylene column packed with 0.4 ml of TRU resin (50-100 μm, Eichrom Technologies, Inc.).Nd was then separated from Sm using a polypropylene column packed with 0.4 ml of Ln resin (100-150 μm, Eichrom Technologies, Inc.) and 0.2 M HCl.All experimental procedures were conducted in a clean room at the GSJ, and all drying processes were conducted in a semi-cylindrical PFA cover purged with nitrogen gas at 4.0 l/min.
The Sr, Pb, and Nd isotopic compositions presented here are the means of 50 measurements using a Neptune multi-collector (MC) -ICP-MS (Thermo Fisher Scientific Inc., U.S.A) at the GSJ.Isotopic ratios were normalized to a 86 Sr/ 88 Sr ratio of 0.1194 and a 146 Nd/ 144 Nd ratio of 0.7219.The measured 87 Sr/ 86 Sr ratio of NBS SRM 987 was 0.710257 ± 0.000006 (2 SD, n = 6), and the 143 Nd/ 144 Nd ratio of JNdi−1 was 0.512086 ± 0.000014 (2 SD, n = 18).The Nd isotopic ratios were normalized to a 143 Nd/ 144 Nd ratio of 0.512115 for JNdi−1 (Tanaka et al. 2000).Pb isotopic ratios were determined using a Tl spike, NIST SRM 997,and

Zircon U-Pb geochronology
Zircon grains for U -Pb dating were separated using heavy liquids, handpicked under a binocular microscope, and mounted together with the TEMORA 2 (417 Ma; Black et al. 2004), FC1 (1099.9Ma; Paces and Miller 1993), and OD−3 (33 Ma; Iwano 2013) zircon standards and SRM NIST 610 glass standard in an epoxy disk.Sample mounts were polished to expose the centre of the zircon grains, and cathodoluminescence (CL) and backscattered electron (BSE) images were taken to examine the internal structures of the zircon grains, including growth zones, fractures, and inclusions, to select the laser spot locations.The CL and BSE images were obtained using a JEOL JSM−6610 SEM at the National Museum of Nature and Science, Japan.The U -Pb-Th isotopic dating was conducted using an NWR213 LA system connected to an Agilent 7700× ICP -MS at the National Museum of Nature and Science, Tsukuba, Japan.Experimental conditions, measurement procedures, and data reduction followed those of Tsutsumi et al. (2012).The errors quoted are at the 1σ confidence level.U and Th contents and Pb isotopic ratios ( 207 Pb/ 206 Pb and 208 Pb/ 206 Pb) were calibrated using the NIST 610 glass standard as a reference material. 207Pb/ 238 U ratios were calibrated using the TEMORA 2 zircon standard.The other two zircon standards were measured as secondary standards along with the sample zircon grains.Correction for common Pb on concordia plots was carried out using 208 Pb.Weighted-mean and spot ages in this paper are based on 207 Pb-corrected values.The mean age calculations and concordia plots were produced using Isoplot 3.71 (Ludwig 2008).

Occurrence and petrography
A small granitic body was discovered in the Hiroo Complex of the Nakanogawa Group in the Hidaka Belt.The granitic rocks are tonalites and occur as a < 100-mwide body in the Rakko River, ~4.5 km northwest of its mouth (Figure 1c).The tonalite intrudes the surrounding greenstones (meta-dacites; Figure 2a), and the greenstone has changed to reddish brown in a ~ 30cm-wide zone along the contact (Figure 2b).The tonalite near the contact is porphyritic, and consists of finegrained quartz and plagioclase.The tonalites were subject to strong brittle deformation and have been altered variably along with the surrounding greenstones.Although the direct relationship between the tonalites and sedimentary rocks of the Nakanogawa Group is unclear, the greenstones occur with red chert and limestone as blocks in alternating sandstone and mudstone layers, which suggests that the tonalites and greenstones are an allochthonous block in the Nakanogawa Group.Seven tonalite samples were taken from different parts of the body, as well as three samples of the surrounding greenstones that comprise the allochthonous block.
The tonalites are composed mainly of quartz, plagioclase, and hornblende (Figure 2c,d).A small amount of Fe -Ti oxides are also present.The tonalites are variably brecciated, and they are highly altered in general (Figure 2e).A portion almost free from brecciation shows equigranular to seriate texture (Figure 2c), and the grain size of the major minerals varies between ~ 2.0 and 1.5 mm.The modal composition of this sample (17101301E) can be classified as a tonalite in the IUGS quartz -alkali feldspar -plagioclase diagram (Le Maitre 2002; Figure 3a).Plagioclase crystals are euhedral to subhedral, and weak zoning and Carlsbad twins were observed.The plagioclase has been mostly replaced by fine-grained clay minerals (Figure 2c).Quartz occurs as interstitial crystals, and commonly shows undulatory extinction (Figure 2c).Hornblende crystals are subhedral, and exhibit pleochroism ranging from deep greenish brown to pale brown (Figure 2d).Mineral chemistry of fresh portion is Mg# [Mg/(Mg + Fe) in atomic ratio] = 0.44-0.49and classified as magnesio-hornblende defined by Hawthorne et al. (2012) (Table S2).The hornblende is, however, completely altered to chlorite in many cases.Fe -Ti oxides consist of rounded grains measuring <1.0 mm in diameter.At the contact with surrounding dacites, the grain size of the tonalite decreases substantially, and the tonalite is porphyritic with plagioclase and quartz phenocrysts.
The dacites are typically porphyritic with plagioclase phenocrysts (Figure 2f), although aphyric rock is sometimes found.The plagioclase phenocrysts are euhedral to subhedral (~0.5 mm in size), and albitized or altered to clay minerals.The groundmass is composed of finegrained plagioclase laths and brown clay minerals.
Due to the highly brecciated nature of the tonalites, it was difficult to obtain modal compositions of all the studied samples.Thus, we plotted normative quartz -orthoclase-plagioclase compositions (wt.%) on the IUGS quartzalkali feldspar -plagioclase classification diagram, and compared this with the modal composition (vol.%) of the non-brecciated part of sample 17101301E (Figures 2c and  3a).Given the very limited area in sample 17101301E, the normative compositions are similar to the modal compositions.The obtained normative compositions of the tonalites and dacites plot in the tonalite field in the IUGS classification diagram.In terms of normative anorthitealbite-orthoclase compositions, following the classification of Barker (1979), all samples plot in the field for trondhjemite (Figure 3b).In this paper, we follow the IUGS classification scheme, and refer to these rocks as tonalites.

Major-and trace-element compositions
The whole-rock major-and trace-element compositions of the studied samples are listed in Table S1.The tonalites have SiO 2 contents of 71.0-73.7 wt.%, and the dacites have SiO 2 contents of 70.1-71.7 wt.% (Table S1).Although the SiO 2 content of the dacites is slightly lower than that of the tonalites, the two partly overlap.On a total alkali -SiO 2 (TAS) diagram, the samples have similar compositions and plot in the low-K and daciterhyolite fields (Figure 4a,b).Pre-Eocene granitic rocks (granitic rocks from the Hidaka Magmatic Zone; Jahn et al. 2014), altered dacites (appearing to be greenstone) from central to eastern Hokkaido (Daimaruyama greenstones; Yamasaki and Nanayama 2018), silicic rocks from mid-ocean ridges and oceanic plateaus and seamounts (PetDB; Lehnert et al. 2000), trondhjemites -tonalites from the Oman ophiolite (Lippard et al. 1986;Rollinson 2009) and Elder Creek ophiolite, California (Shervais 2008), and adakites and Archaean tonalite -trondhjemite-granodiorite (TTG) suites (Martin et al. 2005) are also plotted on Figure 4 for comparison.Our samples have different K 2 O contents from the rocks from the Hidaka Magmatic Zone and the Daimaruyama greenstones from the Tokoro Belt in eastern Hokkaido (Figure 4a).Such extremely low K 2 O contents are typically found in tonalites and trondhjemites (hereafter, collectively referred to as plagiogranites) in ophiolites (e.g.Coleman and Donato 1979; Figure 4a).Archaean TTG suites are also characterized by relatively low K 2 O contents at these SiO 2 contents (Figure 4a,b; Martin et al. 2005).
On Harker diagrams (Figure 4a,c-j), the contents of all elements in our samples decrease with increasing SiO 2 content, except for Na 2 O, which shows no clear trend.The compositions of the dacite samples differ from those of the tonalites in many respects, but the overall compositional characteristics of the dacites and tonalites are similar to those of ophiolitic plagiogranites and Archaean TTG suites (Figure 4).In particular, our samples and some ophiolitic plagiogranites are characterized by higher Na 2 O contents than samples from the Hidaka and Tokoro belts at the same SiO 2 contents.The alumina saturation index [A/CNK = molar Al 2 O 3 /(CaO + Na 2 O + K 2 O)] values of our samples are mostly 1.03-1.08(Table S1).
The tonalites have chondrite-normalized whole-rock La (La [N] ) contents of 34.Yb [N] contents of 14.4-36.5, and chondrite-normalized rare earth element (REE) patterns show weak enrichment in light REEs (LREEs; La/ Yb [N] = 1.23-2.36;Figure 5a).The REE patterns of the dacite completely overlap with those of the tonalites.The N-MORB-normalized whole-rock trace-element patterns of the tonalites and dacites slope gently up to the left, with clear negative Ti anomalies and weak Nb and Ta anomalies (Figure 5b).Both positive and negative Sr anomalies are observed, depending on the sample.The REE patterns of our samples are similar to those of ophiolitic plagiogranites, but have lower LREE contents than granites from the Hidaka Magmatic Zone and dacitesandesites from the Daimaruyama greenstones (Figure 5a,  c).The compositions of the granites from the Hidaka Magmatic Zone are similar to upper continental crust (Rudnick and Gao 2003).The compositions of the studied samples are within the range of silicic rocks from oceanic settings, although the latter have steeper LREE-enriched patterns than our samples (Figure 5c).The REE patterns of Archaean TTG rocks are clearly different from our samples (Figure 5a).Similarly, the trace-element patterns of the studied samples also resemble those of ophiolitic plagiogranites, except for the clear negative Hf anomalies in samples from Elder Creek ophiolite and low contents of highly mobile Rb in the ophiolitic plagiogranites (Figure 5b).The trace-element patterns of the samples from the Hidaka Belt, Daimaruyama greenstones, and silicic rocks  (2002) and Barker (1979), respectively.In (a), the normative quartz -orthoclase-plagioclase (wt.%) compositions of the tonalites and dacites were plotted on the IUGS quartz -alkali feldspar -plagioclase classification diagram and compared with the modal composition of sample 17101301E (Figure 2c).
On Nb versus Y and Rb versus Nb + Y discrimination diagrams, the studied samples plot in the field of volcanic arc granite (VAG; Figure 5e,f).Silicic rocks from midocean ridges also plot in the VAG field.On a Y versus Sr/Y discrimination diagram, the studied samples plot in the island arc andesite -dacite-rhyolite field (Figure 5g), and their compositions cannot be distinguished from ophiolitic plagiogranites, silicic rocks from oceanic settings, rocks from the Hidaka Magmatic Zone, or the Daimaruyama greenstones.

Sr-Nd-Pb isotopic compositions
No isotopic data have been reported for coeval felsic rocks in the subduction complex along the palaeo-Japan arc to date.The studied tonalites are either differentiated products of basaltic magmatic activity on the Izanagi Plate or of partial melting that was caused by the magmatism.Therefore, the characteristics of volcanic rocks within the Pacific Ocean should provide clues regarding the origin of the tonalites (i.e.parent magma or protolith).On an initial εNd (t) versus 87 Sr/ 86 Sr (t) diagram (Figure 6a), most of the studied samples lie on the high-87 Sr/ 86 Sr side of the MORB field, suggestive of the involvement of seawater.The εNd (t) (+5.2 to + 7.5) is within the range of Pacific MORB and partly overlaps the values for the Shatsky Rise samples (Figure 6a 6b,c), the studied samples (apart from one sample) show a depleted nature similar to that of the Shatsky Rise samples.Alternatively, it has been suggested that Ontong Java, Hikurangi, and Manihiki plateaus may have once been part of a single LIP -the Ontong Java Nui (OJN) -based on the similarity in age and geochemistry of lavas from those plateaus (e.g.Castillo 2004;Taylor 2006).Basalts of the OJN have a wide range of geochemical compositions, and the compositional fields for six types (after Sano et al. 2020) are shown in Figure 6 for reference; the Sr -Nd-Pb isotopic features of the studied samples broadly resemble those of rocks from the OJN.The Sr -Nd-Pb isotopic data for the studied samples indicate that the plagiogranite in the study area has an intermediate composition between the Shatsky Rise and OJN basalts.

Zircon U-Pb dating
Zircon grains for U -Pb dating were separated from sample 17101301B.CL images of representative zircon grains are presented in Figure 7a.The weighted-mean ages and spot ages reported below and in Figure 7 are 206 Pb/ 238 U ages.Most zircon grains have homogeneous cores and bright rims in CL images (Figure 7a), and oscillatory zoning is observed in the rims of some grains.Analytical results are listed in Table S4.A total of 28 spots on 28 grains were analysed, 9 of which were rejected because of irregular signals caused by ablation through the entire zircon grain and extremely low U contents.The weighted-mean age of the main zircon population is 159.1 ± 1.6 Ma (18 spots, MSWD = 1.4, Figure 7c), with a concordant age of 158.5 ± 1.6 Ma (95% confidence interval, MSWD = 0.66, Figure 7b).

Zircon trace-element compositions
The trace-element compositions of zircons from tonalite sample 17101301B are listed in Table S2.On the U/ Yb versus Nb/Yb diagram (Figure 8a), zircons from the studied sample plot in the fields for mid-ocean ridgetype and ocean island-type zircons proposed by Grimes et al. (2015); zircons from basalt in the Mikabu Belt plot in a similar area.On Sc/Yb versus Nb/Yb and U/Yb versus Sc/Yb diagrams, some zircons from the studied sample plot in the mid-ocean ridge-type and ocean island-type fields, but many plot in the field of continental-arc-type zircons owing to their elevated Sc contents relative to Yb (Figure 8b,c).Chondrite-normalized REE patterns of zircons from the studied sample show typical oceanic-crust-type features and compositions (La [N] = 0.04-0.86,Yb [N] = 1790-17,582; Figure 8d).Similarly, on U/Yb versus Y and U/Yb versus Hf diagrams, zircons from the studied sample and from basalt of the Mikabu Belt plot in the field of oceaniccrust zircons (Figure 8e,f) proposed by Aoki et al. (2019).On a diagram of U/Yb versus Nb/Yb, zircons from the studied tonalite sample and from basalt of the Mikabu Belt plot in the field of mid-ocean ridgetype zircons (Figure 8g), whereas zircons from the studied tonalites plot in the field of continental-arctype zircons owing to their elevated Sc contents trondhjemites and tonalites from the Elder Creek ophiolite, California.(b,d) Granitic rocks and greenstones (basalt -dacite) from Central and Eastern Hokkaido, with the compositional range of our samples.(e) Sr/Y -Y diagram.Adakite and Archaean high-Al tonalite -trondhjemite-granodiorite (high-Al TTG) and island arc andesite -dacite-rhyolite (ADR) fields are from Defant and Drummond (1990).Symbols in panels (e)-(g) are the same as in Figure 4.
relative to Yb (Figure 8h), differentiating them from the Mikabu Belt basalt.

Modification of whole-rock geochemistry
As mentioned above, the studied samples have been altered, and brittle fractures were observed in some samples.Accordingly, it is necessary to take account of the effect of post-magmatic alteration on the geochemistry of the studied samples, given that such alteration can modify the primary magmatic composition to various extents.If the alteration criterion of Polat and Hofmann (2003) is applied (LOI <6 wt.%), then none of the studied samples have been strongly hydrated or carbonated (Table S1).However, whether the geochemical features are petrologically meaningful should be evaluated more carefully.A summary of the examination is presented below and a detailed discussion is given in the supplementary text.Among the incompatible elements, HFSEs are generally considered to be immobile during alteration, with Ti and Zr being typical examples.Therefore, we examined the correlation between Ti contents and major-element oxides of tonalites (Fig. S1).For correlations with Ti contents, SiO 2 , Fe 2 O 3 , MnO, and P 2 O 5 show strong correlation coefficient (>0.75) and such correlation between the major element of interest and the immobile element (here, Ti) on bivariate diagrams is interpreted as indicating negligible alteration (Polat and Hofmann 2003).For Al 2 O 3 , CaO, Na 2 O, and K 2 O, the spread in data is wide and the correlations are less clear, but all appear to show a broad trend of increasing element content with decreasing Ti content (for Al 2 O 3 , Na 2 O, and K 2 O) or vise versa (for CaO) (Fig. S1).These elements show weak to moderate correlations (0.10-0.69) but the correlation coefficients are reduced by one or two data points that deviate markedly from the main trend (Fig. S1).The unclear correlations among these elements is considered to reflect substantial elemental modification as a result of plagioclase alteration.Although we cannot necessarily conclude that Al 2 O 3 , CaO, Na 2 O, and K 2 O show a 'true' magmatic trend, the relationship between SiO 2 and Ti suggests that bivariate diagrams (Figure 4) can be used to infer magmatic characteristics if one or two highly disturbed samples are excluded.In addition, the normative compositions calculated from the wholerock chemical compositions are generally consistent with the modal compositions, and any elemental modification caused by alteration is considered to be limited to the small differences shown in Figure 3a.
For trace-element compositions, as with the major elements, the possible influence of alteration on trace-element contents can be at least partly evaluated by element variation patterns (Figure 5a,b).There are large variations in the contents of Rb, Ba, and U, which are relatively mobile during alteration, and a wide compositional range is observed for Sr, which is assumed to be related to plagioclase alteration (Figure 5b).However, even taking into account the variations in their trace-element contents, the tonalites and dacites show a coherent traceelement pattern overall (Figure 5a,b).It is well known that REEs are generally insoluble in aqueous fluids and that a small but systematic variation of the ionic radii with respect to the atomic number leads to a systematic behaviour during the magmatic processes.Thus, the smooth and coherent REE patterns in the tonalites and dacites suggest the absence of any serious modification of these elements.As a check on reliability, equilibrium melt compositions calculated from hornblende REE compositions were compared with whole-rock REE compositions.The calculated equilibrium melt compositions agree well with whole rock REE compositions of tonalites and dacites in abundances (contents) and patterns within partition coefficient uncertainties (Fig. S2).
The above examination implies that the rock has not been altered to lose its primary characteristics in wholerock major-element chemistry and REE composition is generally preserved as a system.Therefore, in this paper, although the absolute values of the analytical results do not necessarily indicate the values of unaltered rock, compositional relationships that show petrologically meaningful trends or values that lie within a narrow compositional range are generally interpreted as indicating primary magmatic characteristics.

Origin and petrogenesis of the tonalites and dacites
The subduction complex is typically composed of terrigenous material and exotic blocks that are accreted by the tectonic transfer of material from the subducting oceanic plate to the overlying plate (e.g.Kusky et al. 2013;Wakita 2015).Therefore, the studied tonalite and dacite block potentially originated from either the subducting oceanic plate or the arc crust of the overlying plate.Investigation of the source materials can distinguish whether or not the tonalite and dacite magmas were derived from arc crust.Alternatively, if oceanic materials are proposed as the source materials, the tonalite and dacite magma could be derived from underplated oceanic materials such as amphibolite in an arc crust, or from the oceanic crustal component itself.In this case, the composition of source materials may be indistinguishable petrologically and geochemically.The critical clue to this distinction is therefore the pressure conditions of magma formation.The formation of felsic magma in the oceanic crust is restricted to low pressure conditions.The oceanic crustal component may also contain intra-oceanic arc rocks (e.g.Yamasaki and Nanayama 2018).This type of rock can be distinguished by its geochemical characteristics rather than by its source material.These possibilities are discussed in more detail below.

source of the tonalites and dacites: oceanic versus arc setting
The ca. 46 Ma granitic rocks in the Hidaka Magmatic Zone are interpreted to be near-trench intrusions formed during subduction of the Izanagi -Pacific Ridge (Yamasaki et al. 2021), and the trace-element patterns of these rocks resemble those of upper continental crust (Figure 5; Rudnick and Gao 2003).The whole-rock geochemistry of the studied samples is clearly different from that of rocks from the Hidaka Magmatic Zone (Figures 4 and 5).This difference suggests that the studied tonalities and dacites were not co-magmatic with the Hidaka rocks and are unlikely to be of continental-arc origin.This is also supported by the large age difference between them.As mentioned above, even silicic rocks from oceanic settings plot in the field of VAG in discrimination diagrams (Figure 5e,f).Thus, the discrimination diagrams do not provide robust evidence for an arc origin.
The zircon U -Pb age of the studied sample (159.1 ± 1.6 Ma) is much older than the depositional age of the Nakanogawa Group (64.4 ± 1.0 to 48.8 ± 0.4 Ma; Nanayama et al. 2019), supporting the idea that the tonalite and dacite are a large block within the Nakanogawa Group.The Hidaka Belt is located to the east (on the palaeo-seaward side) of the Sorachi -Yezo Belt that formed during the Late Jurassic -Early Cretaceous (Figure 1b; Ueda 2016).One of the most likely candidates corresponding to the studied tonalites and dacites is the Oku -Niikap Complex in the Sorachi -Yezo Belt.The Oku -Niikap Complex is a remnant arc that formed on the Izanagi Plate and is overlain by the earliest Cretaceous (~145-130 Ma) radiolarites that mark the minimum age of arc extinction (Ueda and Miyashita 2005).However, this remnant arc has been reconstructed from the existence of arc-like mafic volcanic rocks, including boninitic rocks, and no felsic plutonic rocks have been reported.In addition, the Poroshiri ophiolite, with a zircon U -Pb age of 96.7 ± 2.6 Ma (Kizaki 2000), is found on the eastern edge of the Sorachi -Yezo Belt, at the boundary between the Hidaka and Sorachi -Yezo belts (Figure 1b).There have been no reports of blocks or sediment derived from the Poroshiri ophiolite in the Nakanogawa Group.In addition, only two 160-150 Ma zircons (out of a total of 240 zircons) have been reported from sedimentary rocks of the Nakanogawa Group (Nanayama et al. 2018(Nanayama et al. , 2019)); therefore, it is unlikely that the tonalite -dacite block in this study was derived from the Sorachi -Yezo Belt to the subduction complex of the Hidaka Belt.If the studied tonalites and dacites had been derived from an immature arc, substantial amounts of arc-derived fragments and zircons should be present in the subduction complex of the Hidaka Belt.Arc-like rocks are scarcely represented in the subduction complex, which means that it is geologically difficult to infer the existence of a substantial amount of crystalline crust, that is, the main body of an intra-oceanic arc.Nanayama (1992) and Nanayama et al. (1993) suggested that greenstone blocks within the Hiroo Complex mélange in the southern Hidaka Belt are allochthonous blocks derived from the subduction complex of the Tokoro Belt in Eastern Hokkaido (Figure 1b,c).Although the Nikoro Group in the Tokoro Belt was thought to be an accreted seamount formed during the Cretaceous (e.g.Sakakibara et al. 1986), a new zircon U -Pb age of 151.6 ± 1.8 Ma for an oceanic trachyte in the Nikoro Group has recently been reported (Nara et al. 2019).This age is very close to that of our samples, which suggests that the tonalite -dacite block was derived from the Nikoro Group (Figure 1b).The Daimaruyama greenstones is one such allochtonous block and have a dacitic composition with a typical arc signature and are interpreted to have formed in an immature intra-oceanic arc within the Izanagi Plate (Yamasaki and Nanayama 2018).The Daimaruyama Greenstone is located very close to the studied tonalite -dacite body (Figure 1c).However, the alkaline nature of the Daimaruyama greenstone (enriched in total alkalis, TiO 2 , MnO, and P 2 O 5 , and showing LREE-enriched REE patterns) is inconsistent with the geochemical characteristics of the studied tonalites and dacites (Figures 4 and 5).In addition, the Daimaruyama greenstone formed between 125-100 and ca.60 Ma (Yamasaki and Nanayama 2018), and this age range is substantially younger than the zircon U -Pb age of the studied tonalites.These lines of evidence suggest that the studied tonalities and dacites were not co-magmatic with the Daimaruyama greenstones.Furthermore, the discordant geochemical features imply that the studied tonalites and dacites were unlikely to have originated in an intra-oceanic arc of a different age.
The trace-element compositions of zircons may provide information on tectonic setting.Grimes et al. (2015) proposed a classification of tectono-magmatic provenance based on a compilation of over 5300 analyses of trace elements in magmatic zircons, emphasizing that the U/Yb ratios of zircons varied markedly between continental and oceanic sources.From this perspective, zircons from the studied samples plot within the fields of MOR-type and OI-type zircons and are clearly of oceanic origin (Figure 8a).However, the trace-element compositions of zircons from the tonalite sample are characterized by an elevated Sc content and a high Sc/ Yb ratio (Table S2).As such, the Sc/Yb -Nb/Yb relationships of the studied zircons show continental-arc affinities (Figure 8b).During open-system fractionation of ferro-magnesian minerals, ilmenite, and titanite, Sc would be depleted, driving Sc/Yb ratios down (Grimes et al. 2015).According to Grimes et al. (2015), whole-rock compositions from various tectono-magmatic settings indicate that the Sc/Yb ratio of primitive melts is not particularly sensitive to source, and arc zircons typically have elevated Sc/Yb ratios on account of the enriched Sc and depleted Yb relative to zircons from oceanic crust.The elevated Sc/Yb ratios of the studied zircons are caused mainly by the high Sc contents, and no significant depletion of Yb is observed.Therefore, the high Sc/ Yb ratio of the studied zircons may not imply a continental-arc origin, as it may be caused by significant melting of ferro-magnesian minerals (particularly clinopyroxene) and Fe -Ti oxides owing to the high degree of melting of the source rocks.The similarity and consistency between the chondrite-normalized REE patterns of the studied zircons and zircons from the oceanic crust support an oceanic origin for the studied zircons (Figure 8d).Similar discrimination diagrams using zircon trace elements have also been proposed by Aoki et al. (2019) (Figure 8d-h).Diagrams of U/Yb -Y, U/Yb -Hf, and U/Yb -Nb/Yb (Figure 8e-g) all suggest an oceanic crustal origin.Although zircons from the studied tonalites plot in the continental/oceanic arc-type field in an Sc/Yb -Nb/Yb diagram (Figure 8h), this reflects the high content of Sc.
The whole-rock geochemistry of the studied samples differs from that of rocks in the Hidaka Belt and the Daimaruyama greenstones, and the geochemical difference suggests that the formation mechanism of the studied samples was different from subduction zone magmatism in either the palaeo-Japan (i.e.continental) or immature intra-oceanic arcs.In addition, the trace element geochemistry of the zircon is essentially consistent with an oceanic crustal origin.The above discussions suggest that the studied tonalite and andesite were not derived from arc-like crustal materials, but were derived from some kind of oceanic crustal materials.This means that, at this stage, either accreted oceanic materials or oceanic crustal component could be the origin of the studied tonalite and dacite magmas.

Petrogenesis of the tonalites and dacites
The whole-rock major-element, trace-element, and isotopic compositions of the tonalites and dacites closely resemble each other (Figures 4-6).This strongly suggests that the tonalites and dacites are co-magmatic, and correspond to plutonic and effusive facies, respectively, of a common parental magma.These relationships imply that the dacites were extruded slightly prior to the intrusion of the tonalities.Elevated initial Sr isotopic ratios (Table S4 and Figure 6a) suggest that the studied samples were altered by seawater, or formed where seawater was present.In either case, the parental magma of the studied samples was likely generated in an oceanic environment.
The overall whole-rock geochemistry of the studied samples strongly resembles that of ophiolitic plagiogranites, as represented by the Oman and Elder Creek ophiolites (Figures 4 and 5).Although the geochemical definition of ophiolite and ocean-floor plagiogranites (felsic rocks from mid-ocean ridges, oceanic plateaus, and seamounts in this study) is currently debated owing to the compositional diversity of recently acquired data (e.g.Koepke et al. 2007), plagiogranites from representative ophiolites typically show very low K 2 O contents in accordance with the traditional definition of plagiogranite (e.g.Coleman and Donato 1979; Figure 4a).The studied tonalites are similarly characterized by very low K 2 O contents (Figure 4).As mentioned above, the K 2 O contents of the studied samples may have been affected by alteration; therefore, the observed low K 2 O may not be a primary magmatic feature.However, ophiolitic plagiogranites are also commonly highly altered and in many cases contain secondary epidote, chlorite, actinolite, and albite (e.g.Coleman and Donato 1979;Koepke et al. 2004).Even if not igneous in origin, the similarity in K 2 O contents could reflect similar formation processes.
The tonalites studied here are also characterized by relatively flat chondrite-normalized REE patterns (mean La/Yb [N] = 2.3; Table S1), which are similar to those of the plagiogranites in the Oman ophiolite (axis tonalite, mean La/Yb [N] = 0.8; Pallister and Knight 1981;Lippard et al. 1986;Rollinson 2009) and Elder Creek ophiolite (mean La/Yb [N] = 2.2; Shervais 2008)(Figures 4 and 5).These geochemical features suggest that the tonalities in the study area formed in an ophiolitic setting, and represent either mid-oceanic, supra-subduction zone oceanic, or forearc crust.In any case, intrusion in a submarine environment is likely because the dacites are associated with cherts and limestones.
Two major models have been proposed for the generation of oceanic and ophiolitic plagiogranites: (1) differentiation of a parental MORB, and (2) partial melting of oceanic crust (e.g.Koepke et al. 2007, and references therein).Although many ophiolites have now been shown to have formed in a supra-subduction zone setting, the composition of basaltic rocks in such ophiolites is generally similar to MORBs (e.g.Pearce 2003).It is highly likely that both processes can occur in MOR environments and that both are responsible for the formation of plagiogranites in a variety of modern oceanic environments and ophiolites around the world (Morag et al. 2020 and references therein).In the following, the partial melting origin will be examined first.Afterwards, the possibility of a differentiation origin will be mentioned.
Recent studies have suggested that crustal melting involving a free H 2 O phase, referred to as water-fluxed melting, is an important process for the generation of felsic melt (e.g.Weinberg and Hasalová 2015).The presence of a water-rich fluid phase lowers the solidus temperature so that melting can take place under amphibolite-facies conditions, with the potential to produce voluminous melts (e.g.Weinberg and Hasalová 2015, and references therein).Water-saturated melting experiments conducted on basalt and basaltic andesite at 100, 300, and 690 MPa (Beard and Lofgren 1991), and on samples of sheeted dykes with MORB compositions from the Oman ophiolite at 100 MPa (France et al. 2010), produced dacitic to rhyolitic (SiO 2 = 63-80 wt.%) melts with relatively high Al 2 O 3 (12-22 wt.%) and Na 2 O (2-7 wt.%) contents, and low K 2 O (1-2 wt.%) contents at 800-950°C.Beard and Lofgren (1991) suggested that whilst dehydration melting of amphibolites can generate low-K silicic rocks typical of magmatic arcs, water-saturated melting cannot.In contrast, the compositions of water-saturated partial melts are similar to the whole-rock compositions of the tonalites and dacites, except for large differences in their Al 2 O 3 contents (Figure 9a).Al 2 O 3 and CaO contents of the experimental melts, particularly for the relatively high-pressure experiments (300 and 690 MPa), do not match those of silicic rocks from oceanic settings, although the compositions of the 100 MPa experimental products of Beard and Lofgren (1991) partly overlap with those of the studied samples (Figure 9a).
and Hawaii (ocean-island [OI]-type), and continental-arc (CA-type) zircons proposed by Grimes et al. (2015).Coloured fields in panels e -g and thick grey lines in panel h are after Aoki et al. (2019).(d) Chondrite-normalized REE patterns for the zircons from our samples and basalts from the Mikabu Belt with the compositional range of zircons from igneous rocks in modern oceanic crust.Compositions of zircons from basalt from the Mikabu Belt and the compositional range of oceanic crust are after Sawada et al. (2019).CN denotes chondrite-normalized; chondrite values are after Sun and McDonough (1989).The normative composition of the samples plot on the quartz -albite side of 100 MPa cotectic line of the quartz -albite-orthoclase ternary haplogranitic system, suggesting high-degree melting at ~ 100 MPa (Figure 9b).The compositional variation of the samples is similar to the isobaric differentiation trend at 100 MPa reported by Tuttle and Bowen (1958).The compositional data plotting to the Or-poor side relative to the eutectic point is probably the result of a combination of a high degree of melting and the effects of source rock composition.Beard and Lofgren (1991) suggested that dilution effects related to the high percentage of melt in watersaturated (water-fluxed) melting tend to lower K 2 O in the melt.In addition, variations in the Na/K ratios of the parental materials have a direct control on the Na/K ratios of the melts (Helz 1976;Beard and Lofgren 1991).Basaltic rocks in the reaction zones of mid-ocean ridge hydrothermal systems are extremely depleted in K 2 O relative to less-altered basaltic rocks because of extensive water -rock interaction (e.g.Alt 1995).Such a reaction is expected to precede the higher temperature partial melting.The difference between the Na/K ratios of the studied samples and those under experimental conditions, which is caused by the difference in the source materials, is the most likely cause of the shift of the liquid line of descent of the studied samples towards the Or-poor side relative to the experimental results.The similarities among the compositions of the experimental partial melts and the studied samples, and comparison between the normative composition of the samples and experimental granitic systems, suggest that the dacitic melt was produced by partial melting of pre-existing crust, although the possibility that they were generated through differentiation cannot be fully excluded.
The relatively flat whole-rock REE patterns and Sr/Y versus Y relationship of our samples (Figure 5) do not indicate strong partitioning of Y and heavy REEs into garnet during melting; therefore, the major-and traceelement characteristics are consistent with partial melting of amphibolite-facies oceanic crust at relatively low pressures outside the garnet stability field (e.g.<800 MPa; Poli 1993).The crystallization pressure of hornblende in the studied tonalite is not necessarily the pressure at formation, as it represents the final equilibrium pressure in the magma.However, it may be a necessary condition for magma formation at low pressure, as magma formed at low pressure cannot crystallize at high pressure.Therefore, the final equilibrium pressure was estimated from fresh hornblende in sample #17101301I using the Al-in-hornblende geobarometer.Unfortunately, most of the Al-in-hornblende geobarometers are only applicable to granitic rocks with the lowvariance mineral assemblage: amphibole + plagioclase + biotite + quartz + alkali feldspar + ilmenite/titanite + magnetite + apatite, and this type of geobarometer cannot be applied to the studied tonalite.On the other hand, Ridolfi and Renzulli (2012) have used the total Al content of experimental amphiboles from a variety of starting materials, and their formulation does not specify a specific buffering assemblage.The Al-in-hornblende geobarometer proposed by Ridolfi and Renzulli (2012) gives a range of 82.9-99.8MPa from the fresh hornblende in the studied tonalite (Table S2).This result agrees with the pressure (~100 MPa) inferred from experimental phase relations for the formation of the parent melt of the studied tonalite (Figure 9b).
The plagiogranites in the Elder Creek ophiolite are thought to be subduction zone products formed by a transient event associated with the collision and subduction of a ridge (Shervais 2008).Therefore, the similarity between the geochemistry of the studied samples and that of the rocks from the Elder Creek ophiolite probably reflects mainly the similarity of the source materials and formation processes (i.e.partial melting of oceanic crust at low pressure), rather than their geodynamic settings.In addition, the studied tonalites are generally similar to Archaean TTG in terms of major-element geochemistry, regardless of the significant differences in trace-element geochemistry (Figures 4 and 5).Archaean TTG rocks are generally considered to be the products of melting basaltic materials in a subduction zone, based on their geochemical characteristics and compositional similarities to experimental run products (e.g.Rapp et al. 2003;Martin et al. 2005, and references therein).Archaean TTG suites are characterized by steep, enriched LREE patterns (Figure 5a,  c), indicating melt generation in the garnet stability field (e.g.Moyen and Stevens 2006).Based on the above discussion, it is suggested that the general similarity of the studied tonalites, plagiogranites in the Elder Creek ophiolite, and Oman ophiolite, and similarity in the majornormative quartz -plagioclase-orthoclase (Q-Ab-Or) compositions calculated for our samples onto the experimental haplogranite system (after Johannes and Holtz 1996).The dashed grey line shows fractionation during water-saturated isothermal decompression from 200 to 30 MPa at 890°C (after Blundy and Cashman 2001).The solid grey line shows isobaric water-saturated fractionation at 100 MPa (after Fig. 30 Tuttle and Bowen 1958).We used the projection scheme of Blundy and Cashman (2001) to plot An-bearing compositions, and an oxygen fugacity at the fayalite -magnetite-quartz buffer (FMQ)+2 was assumed for the CIPW normative calculations (France et al. 2010), with a corresponding Fe 2+ /all Fe ratio of 0.6.
element chemistry between the studied tonalites and Archaean TTG suites is mainly due to a similarity in source materials (i.e.production of melt from hydrous basaltic materials).The key difference between these rocks is not their tectonic settings, but the pressure conditions of melting.
To test this hypothesis, we conducted batch melting calculations using the composition of basalts from the Tokoro Belt (Yamasaki and Nanayama 2017; sample 83001B) that were the most probable protoliths of the tonalites and dacites based on similarity of age values and tectonic inference of source materials of sediments in the Hidaka Belt (Nanayama 1992;Nanayama et al. 1993), as discussed earlier.The modal compositions from the high-temperature (850-950°C) melt-forming reaction in the water-saturated (i.e.water-fluxed) experiments of Beard and Lofgren (1991), and mineral -melt partition coefficients for hydrous dacite (Bacon and Druitt 1988), were used in our restite unmixing model (White and Chappell 1977).The modelling results revealed that 10%−30% melting of the basalts produces partial melts with compositions similar to the studied tonalites and dacites (Figure 10a).
Finally, the study examines the possibility that tonalites and dacites can be formed from basaltic magma by magmatic differentiation.Here, we have carried out a numerical simulation of magmatic differentiation using the rhyolite-MELTS (Gualda et al. 2012;Ghiorso and Gualda 2016).The initial composition (parental magma composition) using this simulation is the same as for the batch melting calculation.The parameters required for the rhyolite-MELTS calculation are as follows: hydrous conditions with 2 wt.%H 2 O, pressures of 100 MPa, and a fixed redox state as a QFM buffer condition.It should be noted that the parental magma for differentiation in this case should be essentially anhydrous.Therefore, low water content and oxygen fugacity were assumed.The calculated water content of the final residual magma (5.7 wt.% and 6.4 wt.% for equilibrium crystallization and fractional crystallization, respectively) is comparable to the calculated water content of the melt from the hornblende geobarometer (Table S2).Calculations using rhyolite-MELTS showed that hornblende was not present in the crystallized phases of either the equilibrium crystallization or the fractional crystallization.This may mean that the crystallization of hornblende requires open system crystallization at some stage of differentiation.The differentiation path of the fractional crystallization is in agreement with the studied composition of tonalite and dacite in CaO, but Al 2 O 3 and Na 2 O + K 2 O show a large deviation (Figure 10b).It is therefore unlikely, both in terms of crystallization phase and composition of the residual melt, that magmatic differentiation could reproduce the composition of the tonalite and dacite in the study area.
The above results and interpretations suggest that the studied tonalite and dacite most likely formed by partial melting of crustal rocks in either a mid-oceanic or suprasubduction zone setting under low-pressure (outside the garnet stability field) and water-saturated conditions.It is generally thought that the melting of oceanic crust at low pressures can occur only under the influence of an active spreading centre, in either a mid-oceanic or supra-subduction zone setting (e.g.Koepke et al. 2007 and references therein).However, similar conditions can be expected in the region overlying a large, hot mantle plume.One of the most notable submarine hydrothermal vents on the Galápagos Rift, discovered in 1977, is in an active hydrothermal field on a plume-influenced ridge.The presence of active hydrothermal fields in intra-oceanic plate regions (e. g.Hannington et al. 2011) implies that hydrous partial melting of basaltic rocks can occur in oceanic plateaus or seamounts.The nature of hydrothermally altered basalts from the Shatsky Rise and Ontong Java Plateau has been investigated in several studies (Miyoshi et al. 2015;Sano and Nishio 2015).Both of these LIPs presumably formed along or near fast-spreading ridges with a shallow (<6 km) melt lens (Sano and Nishio 2015, and references therein).Some samples from the Shatsky Rise contain abundant pyrite grains that were probably formed by high-temperature fluid circulation (Miyoshi et al. 2015).Furthermore, the remelting of hydrothermally altered basaltic crust has been proposed for Iceland (Bindeman et al. 2012).
Zircon trace-element geochemistry suggests that the studied tonalites were derived from either a mid-oceanic setting or a highly immature intra-oceanic arc (much more immature than the Daimaruyama greenstones) in a supra-subduction zone setting.Although the studied tonalites and dacites lack the geochemical features of a subduction component, such as enrichment in Rb and Ba relative to Th and Nb (e.g.Pearce et al. 2005), these features could be obscured by severe alteration and may not necessarily apply as discriminators for felsic rocks of partial melting origin.Therefore, the distinction between the two settings cannot be made on the basis of chemical compositions alone but should be considered with respect to the contextual geological information.

Geodynamic setting of Late Jurassic felsic magma formation on the Izanagi Plate
Recent global plate motion reconstructions suggest that the Izanagi Plate was being subducted along the eastern margin of Eurasia from 230 to 60 Ma, and that the Izanagi -Pacific ridge, which was oriented subparallel to the trench, was subducted at 60-50 Ma (e.g.Müller et al. 2016).This suggests that the tonalite -dacite block was located on the Izanagi Plate that was eventually subducted completely.As mentioned above, a mid-ocean ridge is the most likely candidate for the production of felsic magma in an oceanic environment.Yamasaki and Nanayama (2017) proposed that the Nikoro Group basaltic and gabbroic rocks, the most likely protoliths of tonalites and dacites, formed at a plume-influenced ridge -ridge-ridge triple junction based on shallow estimates of lithosphere-asthenosphere boundary depth from E-MORB-like whole-rock geochemistry.Sakai et al. (2019) also suggest a plume -ridge interaction model for the origin of the basaltic rocks in the Nikoro Group.In such areas, high heat fluxes are expected to be sufficient for the partial melting of the oceanic crust and the production of felsic magma.However, felsic rocks in a mid-ocean ridge environment commonly occur as finegrained dykes, and metre-size plutonic bodies are very rare (Koepke et al. 2004).This is because the spreading of the oceanic crust inevitably leads to the solidification and cooling of the magma within a short period of time.The inferred existence of a tonalite -dacite plutono -volcanic system and coarse-grained tonalite lithology would require a large magma reservoir to have existed for a prolonged period.
There have also been several reports of Late Jurassic igneous rocks from the Mikabu Belt in southwest Japan (Figure 1a).Sawada et al. (2019) reported a zircon U -Pb age of 154.6 ± 1.6 Ma from a meta-picrite, and Tominaga and Hara (2021) reported a zircon U -Pb age of 157.0 ± 0.9 Ma from trondhjemites in the Mikabu Belt.The similar zircon U -Pb ages of the plagiogranites from the Mikabu Belt and our samples suggest contemporaneous silicic magmatism occurred on the Izanagi Plate.Hydrous partial melting of basaltic rocks produces silicic melts more readily than dehydration melting.However, the amount of silicic rocks in the oceanic crust and ophiolites is generally small [e.g.<0.5% of the ~ 1500m-long Hole 735B in the Southwest Indian Ridge (Dick et al. 2000) and 0.2% of the total Oman ophiolite surface (Nicolas et al. 2000)].Probably because of such reasons, several limited occurrences of plagiogranitic rocks in the Mikabu, Sorachi -Yezo, and Tokoro belts have been documented (e.g.Miyagi 1978;Iwasaki 1984;Uchino et al. 2017;Tominaga and Hara 2021).Miyagi (1978) reported Sr isotope ratios of trondhjemites ( 87 Sr/ 86 Sr 150 Ma = 0.7029-0.7048)that are the same as those of basalts ( 87 Sr/ 86 Sr 150 Ma = 0.7019-0.7045) in the Sorachi -Yezo Belt.Sr isotope ratios of our samples are within the range of those for basalts and trondhjemites in the Sorachi -Yezo Belt.As mentioned above, the zircon trace-element ratios of the studied tonalites are similar to those observed in basaltic rock from the Mikabu Belt (Sawada et al. 2019) and a partial overlap is observed in the many discrimination diagrams (Figure 8a-f).Chondrite-normalized REE patterns of zircons from the studied tonalite show similar patterns to those from basaltic rock in the Mikabu Belt, and the elevated contents of ΣREE in the tonalite samples (2360-23325 ppm for zircons from the Mikabu basalt, and 5077-17372 ppm for the studied zircons) is explained by the more primitive basalt compositions (Figure 8d).Ichiyama et al. (2014) presented petrological and geochemical data for high-Mg volcanic rocks, including picrites, from the Mikabu Belt, and suggested that the picrites were derived from primary magma with MgO contents of >25 wt.% produced by high degrees of partial melting of a depleted peridotitic mantle source with unusually high mantle potential temperatures (<1650-1700° C; Ichiyama et al. 2014).Voluminous melt with such high temperatures could easily cause the partial melting of preexisting oceanic crust above the superplume head.Furthermore, Ichiyama et al. (2014) suggested that the high-Mg rocks in the Mikabu Belt correlate with those in the Sorachi -Yezo Belt.The volcanic rocks from both belts yield unusually high potential temperatures, estimated using the whole-rock MgO contents of lavas and the similar whole-rock major-and trace-element characteristics of high-Mg picrites and komatiites from oceanic and continental LIPs (Ichiyama et al. 2014).
Overall, given the geological conditions described above, the information available at this time suggests a plumerelated origin for the studied tonalites and dacites.Of these, the most likely geodynamic setting that reasonably satisfies the constraints of partial melting of oceanic crustal material at shallow depths and prolonged high temperature conditions as required by petrologically, extensive contemporaneous felsic magmatic activity in the Izanagi Plate and the chemical composition of associated basalts based on geological contrasts, is oceanic LIP with extensive and abundant magma supply over a period of time.

Insights into magmatism on the Izanagi Plate
The formation of a large oceanic plateau or oceanic LIP in the Late Jurassic-Cretaceous Panthalassa Ocean has been discussed by several authors, as mentioned above, and the overall picture remains debated.Here we evaluate these hypotheses with the addition of our new findings above.
The Sorachi -Yezo Belt includes oceanic plateau-type greenstones formed at ca. 150 Ma (Sakakibara et al. 1999), and Kimura et al. (1994) proposed that the Sorachi -Yezo Belt comprises accreted fragments of a missing twin of the Shatsky Rise, based on geological and palaeomagnetic evidence and plate motions.Tatsumi et al. (1998) also showed the geochemical similarity of the rocks from the Mikabu and Sorachi -Yezo belts and the Shatsky Rise.Magnetic lineations show that the Shatsky Rise formed during the Late Jurassic -Early Cretaceous along the trace of a ridge -ridge-ridge triple junction (Nakanishi et al. 1999). 40Ar/ 39 Ar dating of basalts recovered from Ocean Drilling Program (ODP) Site 1213 on the Shatsky Rise gave an age of 144.6 ± 0.8 Ma (Mahoney et al. 2005).Since this age is significantly younger than those of the Mikabu Belt and our samples, the missing twin hypothesis is now somewhat questionable.In fact, whole-rock Sr -Nd isotopic compositions of the studied samples are similar to those of the OJN rather than the Shatsky Rise (Figure 6).Nevertheless, the nearly identical ages of the trondhjemitic rocks from the Mikabu Belt, our samples, and rocks from the Tokoro Belt strongly suggest that they were derived from a single oceanic plateau.Tatsumi et al. (1998) and Sakai et al. (2019) showed that Mikabu and Tokoro metabasalts, as well as most of the OJN and Shatsky Rise basalts have Hawaiian-type, low-Nb/Y ratios at a given Nb/Zr ratio relative to Polynesian-type basalts.However, Tatsumi et al. (1998) proposed that these basalts originally formed in the South Pacific Superplume region, based on the occurrence of some basalts with Polynesian-type Nb/Y -Nb/Zr characteristics in the Shatsky, Ontong Java, and Sorachi (Aniva) regions.Yamasaki and Nanayama (2017) and Sakai et al. (2019) suggest a plume -ridge interaction model for the origin of the basaltic rocks in the Nikoro Group.All of these studies ascribe the formation of the Late Jurassic basaltic rocks to magmatism at a plume-influenced spreading ridge junction.
MORB-like isotopic signatures of the Shatsky Rise have been difficult to explain with respect to magma source (Mahoney et al. 2005).The Shatsky Rise consists of three major edifices, the Tamu, Ori, and Shirshov massifs, which are aligned southwest -northeast.The largest and oldest Tamu Massif shows homogeneous isotopic features, whereas smaller and younger massifs and small seamount chains adjoining the Shatsky Rise show more heterogeneous isotopic features (e.g.Sager et al. 2016;Sano et al. 2020).As a result of recent geochemical studies, the formation of the Shatsky Rise has been explained by the plume head hypothesis; a transition from voluminous and homogeneous plume head eruptions at the main (Tamu) massif to smaller-scale eruptions from the narrower and heterogeneous plume tail further along the plateau (e.g.Sager et al. 2016).The OJN magma shows no MORB-like signature (e.g.Tejada et al. 2004;Mahoney et al. 2005), so formation at or near the ridge -ridge-ridge triple junction seems to be the main cause of the MORB-like isotopic signature of the Shatsky Rise.Alternatively, mixing of the surrounding mantle into the plume source during the ascent of the plume head rather than the plume tail has been suggested by Campbell and Griffiths (1990) and Griffiths and Campbell (1990).The plagiogranite in the study area has a Sr -Nd-Pb isotopic composition that is intermediate between that of the Shatsky Rise and OJN (Figure 6).Considering that the age of the plagiogranite is slightly older than that of the main massif of the Shatsky Rise, it is suggested that the plagiogranite formed during the early stage of the plateau's formation, and that it was derived from the basalt formed at the plume head.This hypothesis is consistent with the fact that coeval basalts in the Shatsky, Ontong Java, and Sorachi (Aniva) regions that have Hawaiian-type characteristics also have partial Polynesian-type Nb -Y-Zr characteristics.Based on the above inferences, it is possible that the products of the same period of igneous activity in the study area are present in the basement of the main massif of the Shatsky Rise as the 'original' Shatsky Rise (Tatsumi et al. 1998), ~30 km below the seafloor.This can reasonably explain the homogeneous MORB-like isotopic signature of samples from the surface of the main massif, which lack any record of processes occurring during the ascent of the original plume head (e.g.Sager et al. 2016).Tominaga and Hara (2021) proposed that the Mikabu basalts were formed on older oceanic crust of the Izanagi Plate, which was located several thousand kilometres from the Pacific -Izanagi-Farallon triple junction, because the accretion age of the Mikabu Plateau was too young (ca.90 Ma) relative to estimates from the geological record (>110 Ma).Although the inconsistency between the accretion age and formation of the Mikabu basalts should not be overlooked, this apparent discrepancy depends on the reconstructed location of the triple junction and trench, and inferred rate of plate motion used in the model of Müller et al. (2016).The difference between the model of Tominaga and Hara (2021) and that of Kimura et al. (1994) lies in the different recognition of the accretion age and the estimation of parameters such as the plate movement velocity.These would require further investigation and data collection.On the other hand, from a completely different perspective, petrological and geochemical studies suggest or are consistent with the formation of large oceanic plateaus at the ridge -ridge-ridge triple junction associated with the superplume, although the geological bodies and methods involved differ, but common conclusions have been reached (Tatsumi et al. 1998;Sakakibara et al. 1999;Ichiyama et al. 2014;Safonova et al. 2016;Yamasaki and Nanayama 2017;Sakai et al. 2019;Sawada et al. 2019).As mentioned above, this model also coherently explains the geochemical properties of the Shatsky Rise and OJN basalts.
Thus, we suggest that the tonalite -dacite in our study area is a part of a Mikabu -Sorachi -Yezo -Tokoro oceanic plateau, which was formed during the Late Jurassic -Early Cretaceous on the Izanagi Plate (Figure 11a).Radiolarian ages of chert and red shale just above the basaltic pillow lava in the Mikabu, Sorachi-Yezo, and Tokoro belts have a common Late Jurassic age (e.g.Matsuoka 1999; Ueda 2016, and references therein).The radiolarian age of the chert associated with the tonalite-dacite block is uncertain, but the presence of limestone in the block supports a long-distance northward drift from the southern oceans.In addition, the Tonin -Aniva Terrain, which is the continuation of the Hidaka Belt in Sakhalin (Figure 1b), contains melange oliststromes with large allochthonous slabs of Jurassic -Lower Cretaceous felsic volcanogenic rocks (Zharov 2005).The details of these allochthonous slabs are uncertain, but they may have been derived from felsic volcanism on the same plateau.
The similar isotopic compositions and ages of the ca.159 Ma plagiogranites in the Hidaka Belt and rocks from the South Pacific Superplume region suggest that a single large oceanic plateau originally formed at the Izanagi -Pacific-Farallon ridge triple junction as a result of upwelling of the South Pacific Superplume during the Late Jurassic -Early Cretaceous.This plume -ridge interaction could also have generated felsic magma through the partial melting of oceanic crust at shallow levels.Half of the 'original' Shatsky Rise (Tatsumi et al. 1998) eventually accreted as separated fragments in the Mikabu and Sorachi -Yezo belts along the palaeo-Japan trench and the Tokoro Belt along the palaeo-Kuril trench (Figure 11b).Consequently, as mentioned by Nanayama (1992) and Nanayama et al. (1993), the 'exotic block' of the plagiogranitic body in this study was derived from the subduction complex of the Tokoro Belt to the Nakanogawa Group in the Hidaka Belt, which was formed in the paleo-Japan trench during the Eocene (Figure 11c).

Conclusions
The discovery and petrological investigation of ca.159 Ma plagiogranitic rocks in the Hidaka Belt suggest that the Nikoro Group, in the accretionary complex of the palaeo-Kuril trench, contains a fragment of a large Late Jurassic oceanic plateau on the Izanagi Plate.The fragment most likely correlates with the oceanic plateaurelated rocks in the Mikabu and Sorachi -Yezo belts, in the accretionary complexes of the palaeo-Japan trench.Because the Sorachi -Yezo Belt continues to Sakhalin Island (e.g.Kimura et al. 1994;Tatsumi et al. 1998;Ueda 2016), our results indicate the existence of a much larger total volume of volcanic rocks than previously thought (10 5 −10 6 km 3 ; Kimura et al. 1994).Furthermore, isotopic compositions of our samples reasonably fit the evolution of Shatsky Rise and support a genetic link between the Shatsky Rise and the Mikabu, Sorachi -Yezo -Tokoro oceanic plateau suggested by previous studies (Kimura et al. 1994;Tatsumi et al. 1998;Ichiyama et al. 2014;Sawada et al. 2019).Insights into geological correlations between the Mikabu, Sorachi -Yezo, and Tokoro belts allows a more precise reconstruction of the magmatism triggered by the Late Jurassic South Pacific Superplume.Although the hypothesis that the Mikabu and Sorachi-Yezo belts are the missing 'twin' of the Shatsky Rise is debated (Kimura et al. 1994;Tatsumi et al. 1998;Ichiyama et al. 2014;Tominaga and Hara 2021), our model predicts that there is still a missing 'triplet' on the Farallon Plate side of the triple junction, which was eventually accreted along the western margin of North America.

Figure 1 .
Figure 1.(a) Simplified tectonic map of Japan and eastern Eurasia.(b) Tectonic map of central Hokkaido, modified after Ueda (2016) and Nanayama et al. (2019).Depositional ages for subduction complexes are after Ueda (2016), and Nanayama et al. (2019) and references therein, and zircon U-Pb ages for Hidaka Magmatic Zone are after Jahn et al. (2014).(c) Geologic map of the Hiroo area, after Nanayama (1992) and Yamasaki and Nanayama (2018).HR1 and HR2 consist mainly of mudstone-dominant turbidite and massive

Figure 2 .
Figure 2. (a,b) Field photographs and (c -f) photomicrographs of the tonalites and dacites in the study area.(a) Outcrop of tonalite and dacite (now greenstone).(b) Close-up of the intrusive boundary between the tonalites and dacites.The dacites are brown near the boundary, suggesting thermal alteration by the tonalitic intrusion.(c) Weakly deformed portion of tonalite under crossed-polarized light, consisting mainly of euhedral, heavily altered plagioclase and interstitial quartz with undulatory extinction.(d) Euhedral to subhedral hornblende under plane-polarized light.(e) Highly brecciated tonalite under crossed-polarized light.(f) Dacite containing euhedral plagioclase phenocrysts under plane-polarized light.Pl: plagioclase, Qtz: quartz, Hbl: hornblende.

Figure 4 .
Figure 4. Major-element geochemistry of our samples and reference rocks.(a) Total alkali -SiO 2 and (b) K 2 O -SiO 2 volcanic classification diagrams, after Le Maitre (2002).(c)-(j) Whole-rock major-oxide Harker diagrams.Error bars indicate one sigma accuracy of the XRF analyses (bottom left of the diagrams).The data referred to as PetDB were downloaded from the PetDB Database (Lehnert et al. 2000; www.earthchem.org/petdb) on 15 September 2020, using the following parameters: tectonic setting = oceanic plateau, seamount, or spreading centre; rock name = rhyolite or granite.

Figure 5 .
Figure 5. Trace-element geochemistry of our samples and reference rocks.(a -d) Whole-rock trace-element patterns of our samples and reference rocks.Panels (a) and (c) are chondrite-normalized and (b) and (d) are normal mid-ocean ridge basalt (N-MORB)normalized.Chondrite and N-MORB values are from Sun and McDonough (1989).(a,b) Tonalites and dacites from the study area, and

Figure 6 .
Figure 6.Age-corrected Sr, Nd, and Pb isotopic compositions of our samples and reference samples.Analytical 2σ error is less than the symbol size.Indian and Pacific MORB fields are after Mahoney et al. (1998).DM: depleted mantle, EM1: Enriched Mantle 1, EM2: Enriched Mantle 2, FOZO: Focus Zone, and HIMU: high U/Pb mantle.Data for the Shatsky Rise and Pacific MORB is from Mahoney et al. (2005) and Heydolph et al. (2014), and compositional data and groups for the Ontong Java, Manihiki, and Hikurangi Plateaus are from Timm et al. (2011), Sano et al. (2020), and references therein).

Figure 7 .
Figure 7. (a) Representative cathodoluminescence images of analysed zircon grains, (b) age distribution plot, and (c) U-Pb concordia diagram for LA -ICP-MS measurements of sample 17101301B.Analysis spots in (a) are marked with white circles and labelled with the spot number from Table 2 and corresponding age.Data in blue in (b) was rejected as a statistical outlier by Isoplot.The errors on the ages are 1σ. 207Pb* and 206 Pb* in panel (c) indicate radiometric 207 Pb and 206 Pb, respectively.Light-blue ellipse denotes concordia age.

Figure 8 .
Figure 8. Zircon trace-element geochemical compositions of the studied samples and reference rocks.(a -c and e -h) Trace-element ratios of the studied samples and basalt samples from the Mikabu Belt with compositional fields for various tectono-magmatic settings.Compositions of zircons from basalts in the Mikabu Belt are after Sawada et al. (2019).For panels a -c, coloured fields are the 95% proportion of kernel density distributions for compiled datasets of mid-ocean ridge (MOR-type), plume-influenced settings of Iceland

Figure 9 .
Figure 9. (a) Comparison between results of hydrous melting experiments and the compositions of our samples.The compositional range for 100 MPa experiments from Beard and Lofgren (1991) is shown as a light-green field surrounded by a dashed grey line.Symbols other than those shown in the legend of panel (a) are the same as in Figure 4 (including the grey field).(b) Projection of

Figure 10 .
Figure 10.(a) Rare earth element patterns for melts obtained in batch melting models of an inferred basaltic source, which are compared with the compositions of the studied tonalites and dacites.See the text for a detailed explanation.F is the degree of melting.The green line is the composition of a basalt sample from the Tokoro Belt (sample 83001B from Yamasaki and Nanayama 2017).Red lines are the modelled partial melts with compositions similar to the studied tonalites and dacites.(b) Comparison of differentiation trend obtained from rhyolite-MELTS simulations with composition of studied tonalites and dacites.See text for detailed discussion.