Magmatic response to the interplay of collisional and accretionary orogenies in the Korean Peninsula: Geochronological, geochemical, and O-Hf isotopic perspectives from Triassic plutons

Phanerozoic internal and peripheral orogens in Northeast Asia converge toward the Korean Peninsula situated between cratonic Asia and the outboard magmatic arc. Widespread Mesozoic plutons in the peninsula provide first-hand information about the magmatic response to the continental and oceanic plate subduction. The present study addresses this issue using comprehensive (n >1100) whole-rock geochemical, zircon U-Pb geochronological, and O-Hf isotopic data obtained from Triassic gabbro-pyroxenitemangerite-monzonite-syenite-granodioritegranite plutons in the central and southern parts of the peninsula. The intrusion of ca. 265–250 Ma calc-alkaline granitoids, including the high-silica adakite, along the outboard (in present coordinates) Yeongnam Massif is coeval with or slightly younger than the Barrovian metamorphism recognized in fold-and-thrust belts surrounding the inboard (northward, present coordinates) Gyeonggi Massif, suggesting a close link between the collisional orogenesis and subduction initiation as commonly documented in Phanerozoic supercontinents. Subsequent Late Triassic plutons emplaced in the Yeongnam Massif are subdivided into the older (232–224 Ma) magnesian and alkalicalcic to calc-alkalic group and the younger (220–217 Ma) ferroan and alkalic to alkalicalcic group temporally intervened by the geochemically arc-like Andong ultramafic complex (222 Ma). Zircon O-Hf isotopic compositions of the older plutons reflect the mixing of metasomatized lithospheric mantle, young (probably Paleozoic) arc crust, and Precambrian basement crust, whereas those of the younger plutons reflect input of the asthenospheric/lithospheric mantle and mafic lower crust. Meanwhile, the Late Triassic (233–224 Ma) potassic plutons in and around the Gyeonggi Massif represent postcollisional magmatism most likely induced by slab breakoff, which may also have been responsible for the shoshonitic magmatism in the Yeongnam Massif. Spatial differences in the age pattern and O-Hf isotopic signature of inherited and synmagmatic zircons from the potassic plutons indicate a selective contribution from an ancient metasomatized lithospheric mantle beneath the North China-like craton and an allochthonous South China-like lithosphere. The formation of the Triassic plutons could be explained by a series of tectonomagmatic events consisting sequentially of the ridge subduction, low-angle subduction, slab breakoff beneath the collisional orogen, tectonic switch to an extension-dominated arc system, and delamination of an overthickened arc lithosphere.


INTRODUCTION
Orogenic belts have formed in response to interior collisional and peripheral accretionary plate convergence since the onset of plate tectonics on Earth. These two end-member types of subduction, with a broad continuum between them, have left their own characteristic geological and geochemical features among the orogens and associated igneous rocks (Bally, 1981;Ernst, 2005Ernst, , 2010Cawood et al., 2009;Collins et al., 2011). Because orogenesis basically reflects the resistance of a buoyant lithosphere to plate underflow, the driving mechanism is more obvious in the case of continental subduction.
Continental collisional [also referred to as "Alpine" or Bally's (1981) "type-A"] orogens, represented by wrinkled mountain belts in the internal part of the assembled (super)continent, mark the termination of a Wilson cycle of ocean opening and closing. The collisional boundaries are characterized by distinct geologic features, such as oceanward verging thrust sheets neighbored by elongated foreland basins, regional Barrovian metamorphism linked to crustal thickening, and exhumation of (ultra)highpressure rocks containing diamond or coesite (e.g., Ernst et al., 1997;Rumble et al., 2003;Song et al., 2014). On the other hand, accretionary [also termed as "Pacific" or Bally's (1981) "type-B"] plate convergence at sites of oceanic subduction has produced subparallel belts consisting of an outboard trench complex, a medial forearc basin, and an inboard arc, each of which has undergone contrasting pressure-temperature conditions (Miyashiro, 1961). The negative buoyancy of the sufficiently aged and cooled oceanic lithosphere provides most of the force required for its sinking and rollback in subduction zones and consequent mantle convection (Stern, 2004, and references therein). This "Western Pacific" type [or "Mariana" (Uyeda and Kanamori, 1979) type] of subduction leads to the development of a retreating arc system, where upper plate extension promotes the formation of arc-flanking basins. Conversely, the advancing or "Andean" [or "Chilean" (Uyeda and Kanamori, 1979) type] arc system induced by the subduction of buoyant oceanic lithosphere, terrane accretion, or plate reorganization results in the development of retroarc fold-andthrust belts (Cawood et al., 2009). The accretionary orogenic system may have switched between phases of advance and retreat, possibly caused by transient flat subduction (Collins, 2002;Beltrando et al., 2007;Li and Li, 2007).
Magmatic responses to collisional and accretionary orogenies differ in terms of their driving mechanism and general geochemical characteristics. Collisional orogeny is typically associated with peraluminous, S-type leucogranites generated by partial melting of the overthickened crust virtually without mantle input (Sylvester, 1998;Barbarin, 1999). It has also been recognized that alkaline igneous rocks and carbonatites are concentrated in some suture zones in Africa, the Kola Peninsula in Russia, the Himalayan Mountains, and in India (Burke et al., 2003;Burke and Khan, 2006;Hou et al., 2006;Leelanandam et al., 2006). The generation of these rocks is considered to have been facilitated by density-and thermally-driven vertical processes such as delamination and slab breakoff. However, collisional convergence does not generally generate a large volume of magma, probably due to the limited supply of volatile components from the descending sialic plate (Ernst, 1999(Ernst, , 2010Davidson et al., 2007). In contrast, continued subduction of oceanic plate commonly produces substantial amounts of juvenile and recycled crust composed mainly of calc-alkaline granitoids. Zircons' Hf isotopes reflect such differences. According to Collins et al. (2011), the range of Hf isotope signatures for the accretionary orogens in the circum-Pacific narrows and trends toward more radiogenic compositions since 550 Ma, whereas the range of signatures from the collisional Phanerozoic orogens in China and the Himalayas becomes broader. These contrasting patterns demonstrate the fundamental difference between oceanic and continental plate subduction in the productivity of juvenile crust.
The generation of arc magma is explained by the melting of the mantle wedge (or of the downgoing slab itself) in response to the addition of fluids from the subducting oceanic lithosphere (Gill, 1981;Peacock et al., 1994;Tatsumi and Eggins, 1995;Tatsumi, 2005;Ernst, 2010;Grove et al., 2012), or in the case of the continental arc, by the influx of underthrusting fertile materials from the retroarc area to the arc root (Ducea, 2001;Ducea and Barton, 2007;DeCelles et al., 2009). The retreating arc system typically displays the trenchward younging trend of igneous activity, whereas magmatism tends to migrate inboard or temporally shuts off under the advancing arc system (Cawood et al., 2009;Ducea et al., 2015).
The Paleozoic-Mesozoic history of the northwestern edge of the circum-Pacific is characterized by the interplay of continental and oceanic plate convergence (Engebretson et al., 1985;Ernst et al., 2007). Phanerozoic internal and peripheral orogens in Northeast Asia converge toward the Korean Peninsula, situated in a tectonic link between cratonic Asia and an outboard magmatic arc. Widespread Mesozoic plutons exposed in the peninsula provide firsthand information about the individual or superimposed roles of collisional and accretionary orogenies in making the magma. The present study addresses this issue using new and published whole-rock geochemical, in situ zircon U-Pb geochronological, and O-Hf isotopic data obtained from gabbro-pyroxenite-mangeritemonzonite-syenite-granodiorite-granite plutons in the central and southern parts of the peninsula that intruded approximately coevally during the Triassic, a period marking the culmination of the collisional and accretionary orogenies. This comprehensive data set (n >1100) bears important implications for deciphering key issues concerning the magmatic response to the continental and oceanic plate subduction, such as the driving mechanism of collision-and arcrelated magmatism, the geodynamic evolution of the arc system, and the specific contributions of mantle and crustal reservoirs to the generated magmas.

GEOLOGICAL BACKGROUND
The Pacific (and the former Panthalassic) Ocean, created at the breakup of Rodinia, has never been closed during its life (Dalziel, 1991;Hoffman, 1991). Prolonged subduction of the (paleo-)Pacific plate, initiated in the latest Neoproterozoic and still ongoing, has left quasicontinuous accretionary orogens along the margin of the circum-Pacific. Opposed accretionary orogens existing on either side of the Pacific Ocean differ substantially in the kinematic framework of the overriding plate, as exemplified in the classic classification scheme of oceanic subduction zones (Uyeda and Kanamori, 1979;Uyeda, 1982). The present-day western Pacific margin has undergone several episodes of tectonic switch between retreating and advancing arc systems since the late Paleozoic (Li et al., 2012).
Eurasia is a tectonic collage of Gondwanan landmasses that have been transported northward since the middle Paleozoic (Sengör et al., 1988;Li and Powell, 2001;Collins, 2003). The successive collisional orogens developed between continental blocks in the interior of Eurasia tend to be progressively younger southeastward from the Paleozoic Caledonides in Europe to the Cenozoic Himalayan orogen between India and Asia. The Qinling-Dabie-Sulu orogen in central-eastern China is in the middle of such a spatiotemporal sequence. Triassic (ca. 240-220 Ma) high-pressure and ultrahighpressure mineral parageneses have been reported from the leading edge of the Yangtze Block in South China (Hacker et al., 2000(Hacker et al., , 2004. Before the Triassic continental collision, the Qinling-Tongbai-Hong'an areas to the west of the Dabie-Sulu Mountains had experienced a series of oceanic subduction and arccontinent collision from the latest Cambrian to the Carboniferous (Wu and Zheng, 2013, and references therein). The tectonic evolution of the curvilinear orogen extending eastward from the Qinling-Dabie-Sulu belt is complicated by severe tectonic and magmatic overprinting and transverse fault activities (Ernst et al., 2007), making it difficult to trace the continuation of this belt into the Korean Peninsula.
Widespread Phanerozoic plutons in Northeast Asia are believed to have been chiefly produced in association with the (paleo-) Pacific plate subduction and, to a lesser extent, with continental collision between the North and South China cratons (Sagong et al., 2005;Yang et al., 2005Yang et al., , 2007aYang et al., , 2007bJahn, 2010;Zhang et al., 2014;Zhu et al., 2014;Cheong and Jo, 2017). Such a presumed connection between plate convergence and magma generation is corroborated by the distribution of the plutons concentrated along the plate boundaries (Fig. 1). The initiation of Phanerozoic magmatism in the Northeast Asian margin is marked by the occurrence of latest Paleozoic (267-256 Ma) calc-alkaline granitoids in southeastern China, southeastern Korea, and southwest Japan (Li et al., 2006;Horie et al., 2010;Yi et al., 2012aYi et al., , 2012bCheong et al., 2013Cheong et al., , 2014. The magmatism then culminated in the Triassic, Jurassic, and Cretaceous periods, with magmatic lulls of ~20-50 m.y. between the flareups (Sagong et al., 2005;Cheong and Kim, 2012;Zhang et al., 2014;Zhu et al., 2014). The tectonic significance of such cyclic igneous activities has been discussed mainly in view of arc system evolution (Sagong et al., 2005;Li, 2007, Choi et al., 2012;Li et al., 2012).
The basement of the Korean Peninsula is composed of three Precambrian (mostly Paleoproterozoic) massifs (Nangrim-or Rangnim, Gyeonggi, and Yeongnam) intervened by Neoproterozoic to Paleozoic fold-and-thrust belts (Imjingang and Okcheon-or Ogcheon) ( Figs. 1 and 2). The southeastern margin of the peninsula forms the Cretaceous Gyeongsang arc system comprising an arc platform and a backarc basin (Chough and Sohn, 2010). The Korean Peninsula has traditionally been considered to be a coherent part of the North China Craton, as reflected by its alternative name, the Sino-Korean Craton. The shared lithologies and metamorphic evolution histories of the three Precambrian massifs suggest that they collectively comprise the Paleoproterozoic (ca. 2.0-1.85 Ga) arc system and a collisional orogen probably stretching from the eastern North China Craton (Cho et al., 2017b). After a long magmatic quiescence from the Paleoproterozoic to the middle Paleozoic, the peninsula experienced another series of tectonothermal events related to the internal continental collision and external oceanic plate subduction.
Many tectonic models have proposed that the Qinling-Dabie-Sulu belt extends into the Korean Peninsula. The tectonic schemes of Cluzel et al. (1991) and Yin and Nie (1993)    to suggest that the Imjingang and Okcheon belts correspond to the geographic position of the suture(s) (Ree et al., 1996;Chough et al., 2000;Ernst et al., 2007). Alternatively, the schematically proposed Hongseong-Imjingang or Hongseong-Odaesan (for locations, see Fig. 2) belts were considered to be the extended collision zone (Oh, 2006(Oh, , 2016Kwon et al., 2009;Oh et al., 2015). More recently, Cho et al. (2017a) suggested that the Imjingang, Okcheon, and Taean-Hongseong belts, collectively referred to as the Gyeonggi Marginal Belt, represent the Neoproterozoic to Paleozoic rift-and arc-related tectonic slivers built upon the North China-like Gyeonggi Massif.
The compilation of sensitive high-resolution ion microprobe (SHRIMP) data has confirmed that the Gyeonggi Massif experienced two distinct tectonothermal events, in Paleoproterozoic and Triassic times (Cho et al., 2017a). The Triassic crustal thickening event resulted in high-pressure metamorphism highlighted by the occurrence of eclogitic amphibolite in the western Gyeonggi Massif (Oh et al., 2005). The SHRIMP zircon dates (ca. 230 Ma) of the eclogite facies metamorphism, reported by Kim et al. (2006), are mostly discordant, and thus may not constrain the timing of peak metamorphism exactly. The post-collisional stage is represented by the Late Triassic gabbro-mangeritemonzonite-syenite-granite suite, occurring commonly as small stocks in the internal and marginal parts of the Gyeonggi Massif ( Fig. 2) (Cheong et al., 2015a, 2016, andreferences therein). This suite is approximately coeval with rare earth element (REE)-enriched carbonatite in Hongcheon, in the central Gyeonggi Massif (for location, see Fig. 2) , and silica-undersaturated alkaline silicate rocks occurring to the north of the Imjingang Belt (Peng et al., 2008;Yang et al., 2010). The Triassic crustal thickening event was followed by a gravitational collapse at ca. 225 Ma to form the Gyeonggi Shear Zone (Kim et al., 2000).
Approximately comparable SHRIMP zircon ages (262-252 Ma) were reported from gabbros and granites recovered from a drill core (~3400-2700 m in depth) in Pohang ) (see Fig. 2), indicating that Permian rocks comprise the upper part of the basement of the Gyeongsang arc. The Paleozoic-Mesozoic transitional arc magmatism in the Yeongdeok-Andong-Cheongsong-Gimcheon area was followed by the intrusion of Late Triassic plutons in the central and southwestern parts of the Yeongnam Massif (Park et al., 2006;Kim et al., 2011b). The latest stage of Triassic magmatism in the Yeongnam Massif is represented by the occurrence of geochemically arc-like (i.e., enriched in large-ion lithophile elements [LILEs], but depleted in high-field-strength elements [HFSEs]) Andong peridotite-pyroxenite complex (Whattam et al., 2011) and the Daegang Atype granite (for locations, see Fig. 2), of which emplacement ages were constrained at 222 Ma and 220 Ma, respectively (Cho et al., 2008;Jeong et al., 2014).
After the Permian-Triassic tectonomagmatic events, the Korean Peninsula was subjected to an accretionary orogenic system. The Jurassic and Cretaceous magmatism migrated inland and trenchward, respectively (Kee et al., 2010;Cheong and Kim, 2012;Choi et al., 2012), perhaps resulting from the tectonic switch between the advancing and retreating arc systems. The Phanerozoic plutons comprise around one-third of the total landmass of the southern Korean Peninsula (Fig. 2). They are subdivided into four spatiotemporal groups that intruded episodically in the Permian-Triassic, the Early Jurassic, the Middle Jurassic, and the Cretaceous to Paleogene (Cheong and Kim, 2012). Triassic, Jurassic, and Cretaceous igneous activities were also identified by detrital and magmatic zircons collected from the North Korean territory .

MATERIALS AND METHODS
This study used new and published wholerock geochemical, zircon U-Pb geochronological, and O-Hf isotopic data for 21 gabbropyroxenite-mangerite-monzonite-syenitegranodiorite-granite samples collected from Triassic plutons in the Gyeonggi (10 samples) and Yeongnam (11 samples) massifs. The sample locations are shown in Figure 2, and their GPS coordinates are presented in Table DR1 1 , with petrographic and mineralogic summaries. Eleven new samples were collected for this study. They were three granites from the western Gyeonggi Massif (Taean, Seosan, and Namyang) and eight monzonite-mangeritesyenite-granodiorite-granite samples from the central Yeongnam Massif (Daegang, Macheon, Hamyang, Hapcheon, Sancheong, and Sangju). The other ten samples were collected previously for geochemical and isotopic studies. Three (monzonite and syenite), two (gabbro and monzonite), and two (mangerite) samples were collected from the western (Gwangcheon and Haemi), central (Yangpyeong), and eastern (Odaesan) Gyeonggi Massif, respectively, by Cheong et al. (2015a). Two orthopyroxenites from the Andong ultramafic complex  and one granodiorite from the Yeongdeok pluton (Yi et al., 2012a) were also analyzed in this study.
Zircon U-Th-Pb isotopic analyses were carried out using a SHRIMP IIe/MC at the Korea Basic Science Institute (KBSI) (Ochang, Korea). Prior to the SHRIMP analysis, cathodoluminescence (CL) and backscattered electron (BSE) images of separated zircon grains were examined using a scanning electron microscope (JEOL JSM-6610LV) at the KBSI. Zircon U-Pb ages were newly determined for four samples from Taean in the Gyeonggi Massif, and Hapcheon and Sancheong in the Yeongnam Massif. In addition, seven samples from the Gyeonggi Massif and four samples from the Yeongnam Massif were dated again in this study considering the relatively high scatter of previous SHRIMP zircon ages (i.e., relative error >1%, mean square weighted deviation [MSWD] >5). Whole-rock major element analyses for eleven samples (three from the Gyeonggi Massif and eight from the Yeongnam Massif) were performed at Pukyong National University (Busan, Korea) and Activation Laboratories Ltd. (Canada) using an X-ray fluorescence spectrometer. Trace element compositions of sodium peroxide-fused whole-rock powders (ten samples from the Gyeonggi Massif and eight samples from the Yeongnam Massif) were determined at Activation Laboratories Ltd. with a quadrupole inductively coupled plasma-mass spectrometer (ICP-MS). Zircon oxygen isotopes were measured for seventeen samples (eight from the Gyeonggi Massif and nine from the Yeongnam Massif) using the Cameca IMS 1280 ion probe at the secondary ion mass spectrometry laboratory of the Institute of Geology and Geophysics, Chinese Academy of Sciences in Beijing, China. Zircon Lu-Yb-Hf isotopes 1 GSA Data Repository item 2018374, Tables DR1-DR4: GPS coordinates and petrographic and mineralogic summaries of whole-rock samples, SHRIMP U-Th-Pb results for zircon, whole-rock major and trace element compositions, and zircon O and Lu-Yb-Hf isotopic compositions, is available at http://www.geosociety.org/datarepository/2018 or by request to editing@geosociety.org.
were measured for nineteen samples (eight from the Gyeonggi Massif and eleven from the Yeongnam Massif) using a Plasma II multicollector ICP-MS (Nu Instruments) equipped with an NWR193-nm ArF Excimer laser ablation system at the KBSI. Laser ablation was targeted on the points selected for the determination of U-Th-Pb-O isotopes or new points within the same CL domains. Details of the zircon analyses are available in Appendix 1.

Zircon Texture and U-Pb Age
Representative CL and BSE images of the zircon grains are shown in Figure 3. The SHRIMP zircon U-Th-Pb data are listed in Table DR2 (see footnote 1). Figure 4 shows the zircon U-Pb isotopic composition of eight samples from the Gyeonggi Massif in concordia diagrams.

Yeongnam Massif Plutons
Zircon U-Pb isotopic compositions of seven samples from the Yeongnam Massif are graphically displayed in Figure 5.

Whole-Rock Chemical Composition
Whole-rock major and trace element compositions of the plutons are listed in Table DR3 (see footnote 1). Some major element classification diagrams are presented in Figure 6 using new data obtained in this study and previous results from the literature (Cheong et al., 2002(Cheong et al., , 2015aOh et al., 2006b;Choi et al., 2009;Williams et al., 2009;Kim et al., 2011bKim et al., , 2011dYi et al., 2012a).
In the classification scheme of Frost et al. (2001), mafic to intermediate (SiO 2 ≤ 60 wt%) samples from the Gyeonggi Massif were invariably magnesian (Fig. 6A). Granite samples from the western parts of the massif (Taean, Seosan, Hongseong, Haemi, and Namyang) were scattered between magnesian and ferroan fields. Of the Yeongnam Massif samples analyzed in this study, only alkali granite and syenite samples from Daegang, Hapcheon, and Sancheong were ferroan. These ferroan samples were relatively young (220-217 Ma) compared with the other Yeongnam samples (232-224 Ma) mostly belonging to the magnesian group. The Gyeonggi Massif samples were plotted predominantly  in alkalic and alkali-calcic fields, whereas the Yeongnam Massif samples straddled the region between alkali-calcic to calc-alkalic fields, except for two syenites from Hapcheon and Sancheong that were plotted distinctly in the alkalic field (Fig. 6B). At the given SiO 2 content, the Gyeonggi samples tended to be more enriched in K than the Yeongnam samples (Fig. 6C). Mafic to intermediate (SiO 2 = 60-45 wt%) samples from the Gyeonggi Massif were mostly plotted in the shoshonite series field, whereas felsic (SiO 2 ≥70 wt%) samples were dispersed from the shoshonite to high-K series fields. The Yeongnam Massif samples straddled the region between the high-K and medium-K series fields, except for the ferroan samples and the Macheon monzonite plotted in the shoshonite series field. The geochemical difference between the Gyeonggi and Yeongnam samples was most sharply distinguished by the K 2 O/Na 2 O ratio (Fig. 6D). The dominantly potassic (K 2 O > Na 2 O, in wt%) and sodic (Na 2 O > K 2 O) signatures of the Gyeonggi and Yeongnam samples were not correlated with their SiO 2 contents. Among the Yeongnam samples analyzed in this work, only the Daegang alkali granite was higher than unity in K 2 O/Na 2 O ratio.  accepted boundary value between I-and S-type granites (Chappell and White, 1992).
Of the samples analyzed in this study, the mafic to intermediate rocks from the Gyeonggi Massif had distinctly higher Ba contents (3040-1950 ppm) than the other samples (≤1300 ppm). At given SiO 2 contents, the Gyeonggi samples tended to have higher Cs, Rb, and Pb content than the Yeongnam samples. Five syenitegranodiorite-granite samples from Hapcheon, Sancheong, Hamyang, and Daegang in the Yeongnam Massif, and Taean in the Gyeonggi Massif were relatively abundant in Y (62-44 ppm) compared with the other samples (<30 ppm). The Zr and Hf concentrations of all samples were strongly and positively correlated (R 2 > 0.95). The Sancheong syenite had the largest abundance of these elements (Zr = 1020 ppm, Hf = 20 ppm). The Nb concentration was the highest in the Taean granite (49 ppm). Overall, both the U and Th contents increased but the V, Cr, and Ni contents decreased progressively with increasing SiO 2 content. The Haemi syenite from the western Gyeonggi Massif was exceptionally high in Th (53 ppm). Figure 7 shows the normalized geochemical patterns of the plutons analyzed in this study. The average composition of the Yeongdeok adakite pluton (Cheong et al., 2002) is also shown for reference. Most Gyeonggi samples displayed LREE-enriched chondrite-normalized patterns (Fig. 7A). The exception was the Taean granite, which displayed the "seagull" chondritenormalized pattern (see Glazner et al., 2008) with a deep Eu anomaly (Eu/Eu* = 0.02). The older (232-224 Ma) Yeongnam Massif samples had LREE-enriched normalized patterns, without conspicuous Eu anomalies, except for the Hamyang granodiorite, which had a high heavy REE content and stronger negative Eu anomaly (Eu/Eu* = 0.39). The younger (220-217 Ma) ferroan Yeongnam Massif samples mimicked the REE pattern of the Hamyang granodiorite. In the normal mid-oceanic ridge basalt-normalized spider diagram, which arranges the elements according to their compatibility with the basaltic liquid, the Gyeonggi and Yeongnam samples exhibited apparent arc affinities characterized by an enrichment in LILEs and relative depletion in HFSEs (Fig. 7B). Peaks for K and Pb, and troughs for Nb, P, and Ti were prominent in most samples. The troughs were particularly distinct in the Taean granite.

Zircon O-Hf Isotopes
The new zircon O and Lu-Yb-Hf isotope data are listed in Table DR4 (see footnote 1). The Gyeonggi Massif plutons are hereafter subdivided into the western and central-eastern groups, considering their substantial differences in inherited zircon age patterns. Cheong et al. (2015a) revealed that these two spatial groups differ profoundly in their whole-rock Nd and zircon Hf isotopic compositions. It is also noted that the western group occurs within the Taean-Hongseong Complex composing the Gyeonggi Marginal Belt proposed by Cho et al. (2017a). The zircon ages and whole-rock geochemical compositions described above allowed us to subdivide the Yeongnam Massif plutons into the older (232-224 Ma) magnesian group and the younger (220-217 Ma) ferroan group. Zircon U-Pb ages and O-Hf isotope data are summarized in Table 1 for the four subgroups using data from the present analyses and the literature Cheong et al., 2015a).

Younger Group of Yeongnam Massif Plutons
Except for dark-CL spots, the zircon δ 18 O values of the Daegang alkali granite were quite consistent (6.04 ± 0.19‰). As shown by zircon grains from the Taean granite in the western Gyeonggi Massif, the uraniferous dark-CL spots had comparatively low δ 18 O values (5.63 and 5.89‰). The ε Hf (t) values varied moderately from 1.1 to -5.3. Zircons from the Hapcheon syenite had the lowest δ 18 O (5.10 ± 0.17‰) and the second highest ε Hf (t) (8.1 ± 1.1) values among the synmagmatic grains from the plutons analyzed in this study. Zircon δ 18 O values of the Sancheong syenite varied from 8.41‰ to 6.50‰. But this sample yielded consistent zircon ε Hf (t) values (3.0 ± 0.6). Combined with data from Jeong et al. (2014), the zircon Hf isotopic compositions of the two pyroxenites from Andong differed slightly from each other (ε Hf (t) = 1.7 ± 1.0 and 3.6 ± 1.2). However, their zircon O isotopic compositions were comparable, yielding an average δ 18 O value of 7.08 ± 0.26‰.

Temporal Correspondence of Collisional Orogeny and Arc Magmatism
After Paleoproterozoic (ca. 2.0-1.85 Ga) accretionary and collisional events, the cratonic part of the Korean Peninsula remained tectonically and magmatically calm until the continental collision between the North and South China cratons, although it should be noted that Neoproterozoic-Paleozoic magmatic and metamorphic events have occurred along the fold-andthrust belts surrounding the Gyeonggi Massif (Gyeonggi Marginal Belt; Cho et al., 2017aCho et al., , 2018. The Phanerozoic orogenies left a strong imprint on pre-existing rocks, and directly or indirectly caused the formation of magma. In the following, the temporal relation between the Phanerozoic orogenic event(s) that affected the Gyeonggi Massif and the surrounding Imjingang and Okcheon belts, and arc magmatism along the Yeongnam Massif is discussed based on our dating results combined with available age data from the literature. The age data for the Permian-Triassic metamorphic and magmatic events are compiled in Table 2.
Metamorphic ages reported from the Imjingang Belt are rare. Ree et al. (1996) reported an earliest Triassic Sm-Nd age (249 ± 31 Ma) using hornblende, garnet, and plagioclase separated from a Neoproterozoic amphibolite comprising the southern part of the belt (the Samgot unit). This age was later confirmed by SHRIMP U-Pb dating of zircon overgrowth rims from a garnet-biotite gneiss (252.9 ± 1.9 Ma; . Recently, Late Triassic magmatic ages were reported from the adjacent Nangrim Massif in the North Korean territory. Peng et al. (2008) presented a zircon crystallization age of 224 ± 4 Ma for a biotite syenite that comprises a gabbro-pyroxenite-syenite complex (the socalled Tokdal Complex) probably stretching from eastern China. A comparable Rb-Sr age (223.3 ± 6.6 Ma) was obtained from phlogopite in a kimberlite associated with the Tokdal Complex . These two age data, together with a monazite Th-Pb age of the Hongcheon carbonatite in the Gyeonggi Massif (232.9 ± 1.6 Ma; Kim et al., 2016), confirm the occurrence of Late Triassic mantle-derived magmatism in central Korea. Cho et al. (2017a) constrained the timing of Triassic crustal thickening event in the Gyeonggi Massif to 245-230 Ma based on a compilation of the SHRIMP U-Pb and Th-Pb ages of accessory minerals. Their new SHRIMP monazite (235-231 Ma) and zircon ages (ca. 226 Ma) from the Mount Cheonggye area near Seoul, Korea, may indicate that the Triassic collisional orogeny was accompanied by rapid cooling and exhumation. However, we note that at least part of zircon rim ages reported previously from the Gyeonggi Massif should be interpreted with caution. Importantly, it is quite difficult to distinguish the age of a regional metamorphic event from that of a local thermal overprint. The influence of the magmatic overprint on the zircon age was typically exemplified in the Yangpyeong area, where Paleoproterozoic basement gneisses yielded a zircon U-Pb age of ca. 235 Ma from overgrowth rims or newly grown grains . This age is indistinguishable from the crystallization ages of synmagmatic zircons from gabbros and monzonites in this area (232-228 Ma; Cheong et al., 2015a; this study), considering the error ranges. This is especially true when the pooled age of zircon cores from the monzonite (238.4 ± 5.0 Ma; Cheong et al., 2015a) is taken to represent the crystallization timing of "antecryst" (see Miller et al., 2007). Moreover, among the samples analyzed by Oh et al. (2015), the cordierite-sillimanite-garnet-biotite gneiss (sample no. YP182F), located closest to the gabbro (<1 km), contained an abundance of newly grown zircon grains. The maximum age of Triassic metamorphism was determined by Cho et al. (2017a) (245 Ma) on the basis of zircon ages from basement rocks in the Odaesan area (Oh et al., 2006a;Cho, 2014). In fact, however, the Paleoproterozoic spinel granulite and migmatitic gneiss in this area yielded highly scattered zircon rim dates ranging from 265 Ma to 230 Ma  (Oh et al., 2006a). The zircon crystallization age of the Odaesan mangerite (232-227 Ma; Jeong et al., 2008;Cheong et al., 2015a;this study) is close to the younger boundary of these rim dates. Oh et al. (2006a, page 570) stated that "it is not possible to distinguish undisturbed analyses from those possibly affected by Pb loss or diffusion of older Pb from the cores into the overgrowths." Although the median values of the zircon rim dates adopted by Oh et al. (2006a) (245 ± 10 and 248 ± 18 Ma) are consistent with the later zircon rim age from unconformably overlain biotite schist (247 ± 6 Ma; Cho, 2014), it is noted that this age was based on data from only three spots, and thus the thermal effect cannot be sufficiently evaluated. The Yangpyeong and Odaesan cases suggest that at least part of the Triassic ages reported from the Gyeonggi Massif represent local, post-collisional thermal events. Alternatively, the magmatic overprint may have induced a partial loss of radiogenic Pb. If this is the case, the zircon rim dates from the Odaesan area (265-230 Ma;Oh et al., 2006a) leave the possibility that the crustal thickening event commenced in the Gyeonggi Massif diachronously (i.e., in the late Permian and in the Middle-Late Triassic).
Regional structures in the metamorphic rocks of the Okcheon Belt are considered to have been governed by large-scale thrust faults. Cluzel et al. (1990) divided the metamorphosed southwestern part of the belt (Okcheon Metamorphic Belt [OMB]) into five thrust-bounded lithotectonic units: the structurally overlain units of Pibanryeong and Chungju and the lower ones of Turungsan, Poeun, and Iwharyeong. The isotopic imprint of metamorphism is relatively weak in the OMB, probably due to the moderate peak temperature conditions (490-630 °C; Cho and Kim, 2005). Adachi et al. (1996) reported an earliest Triassic chemical Th-U-total Pb isochron method (CHIME) age of 251 ± 10 Ma for allanite in muscovite-chlorite-quartz schist in the Turungsan unit (traditionally referred to as the Munjuri Formation). On the other hand, early Permian ages (ca. 290-280 Ma) were obtained by whole-rock Pb isotopic and uraninite CHIME data of black slates in the Poeun unit (Cheong et al., 2003). The early Permian ages were later confirmed by U-Pb isotope data of acid step-leaching experiments conducted on pelitic and quartz-hornblende-garnet schists of the Pibanryeong unit (Kim et al., 2007). Allanite Th-Pb dating for REE ores distributed in the Chungju unit (traditionally referred to as the Kyemyeongsan Formation) revealed the presence of multiple age components in the Late Ordovician (445 ± 8 Ma), Permian to Triassic (ca. 300-220 Ma), and Early Jurassic (199-183 Ma) (Cheong et al., 2015c). The significance of the oldest age is not clear yet, but the latter two ages-albeit scattered-may broadly represent the timing of crustal thickening orogeny and subsequent paleo-Pacific plate subduction. Kim et al. (2013b) also reported a Permian metamorphic age (259.7 ± 3.3 Ma), based on SHRIMP U-Pb analyses of zircon rims from a Neoproterozoic volcanic rock in the Kyemyeongsan area. Cho et al. (2018) attributed the thrusting event between the Pibanryeong and Poeun units to the middle-late Permian (ca. 270-250 Ma) "Ogcheon Orogeny." They distinguished this event from the successive Middle-Late Triassic "Songrim Orogeny" related to the North and South China collision. Further geochronological, petrological, and structural studies are required to confirm this hypothesis.
The Paleozoic-Mesozoic metamorphism in central Korea is broadly synchronous with magmatic and thermal events that occurred in the Yeongnam Massif. The arc magmatism occurred extensively along the margin of the Yeongnam Massif in the earliest Triassic (ca. 250 Ma), as represented by the Yeongdeok adakite (Yi et al., 2012a), and sodic metagranitoids and orthogneisses in the adjacent area Song et al., 2015). The initiation of arc magmatism may be traced back to the middle Permian, considering the SHRIMP zircon ages (266-261 Ma) of diorite-granodiorite xenoliths found in the Yeongdeok pluton and Andong tonalitic gneiss (Yi et al., 2012b;Cheong et al., 2013Cheong et al., , 2014). An even older, middle Paleozoic arc magmatism was suggested from the margin of the Yeongnam Massif on the basis of single age populations of detrital zircons (ca. 330-310 Ma, Kim et al., 2012;ca. 430-370 Ma, Cheong et al., 2015b) as typically displayed by zircons from arc-flanking basins (Cawood et al., 2012). It is noted that many detrital and inherited zircons from the Paleoproterozoic basement gneisses in the northeastern Yeongnam Massif and overlying Cambrian supracrustal rocks (the Myeonsan Formation) have experienced severe Pb loss events (Kim et al., 2013a. Some of them yielded a cluster of lower intercept U-Pb ages at ca. 380 Ma , suggesting a possible link between the Pb loss event and the middle Paleozoic arc magmatism. These zircon data collectively indicate that the margin of the Yeongnam Massif has been activated recurrently during the Paleozoic. Further studies are necessary to understand the linkage between the orogenic event and arc magmatism that occurred in the Paleozoic. In conclusion, as graphically summarized in Figure 8, the middle Permian to earliest Triassic (ca. 265-250 Ma) arc magmatism in the Yeongnam Massif is coeval with or slightly younger than the metamorphic events recorded in the Gyeonggi Massif and the Imjingang and Okcheon belts . This coincidence may suggest that internal orogenic event(s) transferred shortening to the margin of the Yeongnam Massif, although there remains uncertainty about what specific orogeny was primarily responsible for subduction initiation. The temporal correspondence between interior collisional and external accretionary oro genesis was well established by a previous study of Phanerozoic supercontinents (Cawood and Buchan, 2007).

Source Characterization by Zircon Data
The age pattern and isotopic signature of inherited zircons provide first-hand information regarding the magma source deep in the crust (Hawkesworth and Kemp, 2006;Roberts and Spencer, 2015). In this study, we found inherited zircon cores from five granitoid samples (Seosan and Namyang granites from the western Gyeonggi Massif, Yangpyeong monzonite from the central Gyeonggi Massif, and Hamyang granodiorite and Sangju granite from the Yeongnam Massif). Their age patterns are shown in Figure 9, together with those of the inherited zircon cores from the Jurassic granitoids in the Gyeonggi Massif and the Okcheon Belt (Jo et al., 2018). As shown in this figure, the age pattern of the zircon inheritance differs profoundly between the western Gyeonggi Massif and the Yeongnam Massif. The Neoproterozoic (0.83-0.65 Ga) component is predominant in the former with subordinate Paleoproterozoic (2.3-2.1 Ga) and Mesoproterozoic (1.4-1.1 Ga) populations. In contrast, the Orosirian (ca. 1.85 Ga) component is overwhelming in the latter, which is also the case for the Yangpyeong monzonite and Middle Jurassic (177-167 Ma) granitoids in the interior Gyeonggi Massif. Meanwhile, the Early Jurassic  granitoids in the central Okcheon Belt have a comparable inherited zircon age pattern with the Seosan and Namyang granites analyzed in this study. To summarize, whereas the ca. 1.85 Ga zircon inheritance is predominant in the interior Gyeonggi and Yeongnam massifs, the Neoproterozoic inheritance is prominent in the western Gyeonggi Massif and the Okcheon Belt. The former is typically observed in North China, in association with the collision between the eastern and western blocks (Zhao et al., 2002), while the latter is more popular in South China related to the assembly and breakup of the Rodinia supercontinent (Li et al., 2003).
The exotic feature of the protolith of the Early Jurassic granitoids was further highlighted by the significantly low δ 18 O values 4.9 to -0.9‰) of Cryogenian (ca. 750-700 Ma) zircon cores (Jo et al., 2018). Neoproterozoic rocks and zircons from South China, especially from the northern margin of the Yangtze Block, commonly have δ 18 O values substantially lower than the normal mantle range (5.3 ± 0.3‰; Valley et al., 1998), possibly resulting from the hightemperature reaction of their source rocks with glacial water (Zheng et al., 2008, and references therein). Jo et al. (2018) therefore interpreted that the geochronological and oxygen isotopic contrasts of inherited zircon cores observed between the Early and Middle Jurassic granitoids resulted from selective contributions from South and North China-like terranes that had been juxtaposed infracrustally along the margin of the Gyeonggi Massif during the prior collisional orogeny.
The zircon ε Hf (t) and δ 18 O values of the Triassic plutons are plotted against their corresponding crystallization ages in Figure 10. In this figure, O and Hf isotope data for the inherited zircon cores from the Jurassic granitoids (Jo et al., 2018) are also shown for reference. Although the Cryogenian oxygen isotopic feature was not observed in the Neoproterozoic zircon cores from the Triassic Seosan and Namyang granites (δ 18 O = 8.4-6.9‰), zircon Hf isotopes reveal the fundamental difference in the evolutionary history of the crustal source between granitoid rocks in the western Gyeonggi Massif and the Okcheon Belt, and those in the interior Gyeonggi and Yeongnam massifs. As shown in Figure 10B, Neoproterozoic zircon cores in the former granitoid group have chondritic or positive ε Hf (t) values. This range cannot be explained by the evolution of the North China crust, which has a zircon Hf model age typically between 3.4 and 2.7 Ga (Geng et al., 2012;Kim et al., 2014), but is most strongly correlated with ε Hf (t) values of the Neoproterozoic (780-750 Ma) rocks in the northern and western margins of the Yangtze Block in South China (~11-2; Liu et al., 2009;Zheng et al., 2007). Jo et al. (2018) suggested that mixing between this Neoproterozoic component with near chondritic ε Hf (t = Jurassic), and the Archean-Paleoproterozoic component with significantly negative ε Hf (t = Jurassic) resulted in a diverse range of Hf isotopic compositions of the magmatic zircons in the Early Jurassic granitoids. In summary, the inherited zircon data of the Triassic and Jurassic granitoids collectively indicate that the infracrustal basements of the interior Gyeonggi and Yeongnam massifs and the marginal parts of the Gyeonggi Massif (the western Gyeonggi Massif and the Okcheon Belt) were composed of North Chinaand Yangtze-like terranes, respectively. The conspicuous lack of a ca. 1.85 Ga zircon inheritance in the Seosan and Namyang granites suggests that the basement crust beneath the western Gyeonggi Massif consisted of solely Yangtze-like rocks during the Late Triassic.
Oxygen and hafnium isotope data of synmagmatic zircons from the Triassic plutons are shown in Figure 11. The data from uranifer-ous zircon domains in the Taean and Daegang granites were not considered here because the correlation between U contents and δ 18 O values indicates the post-crystallization interaction of metamictized parts with hydrothermal fluids (see Gao et al., 2014). The significantly negative zircon ε Hf (t) values of the Yangpyeong gabbro and the Odaesan mangerite (-18 to -24) make it evident that the chemical fractionation of the lithospheric mantle beneath the interior Gyeonggi Massif occurred in the distant past. This antiquity is also evidenced by a substantially negative ε Nd (t) (~ -26) value of monazite from the Hongcheon carbonatite . The exact mechanism and timing of this event are difficult to ascertain, but broad geochemical arc affinities of the gabbro and mangerite (Fig. 7B) and the Re-Os isotopic study of Lee and Walker (2006) imply that it occurred in the Paleoproterozoic time under an arc environment. Probably, the lithospheric mantle beneath the Nangrim Massif also formed in geologically ancient time, considering the significantly negative whole-rock ε Nd (t) values of the Tokdal Complex (-14 to -20;Peng et al., 2008). The metasomatized lithospheric mantle origin of the Yangpyeong gabbro and Odaesan mangerite is consistent with their zircon δ 18 O values (7.49-6.94‰) being far from the normal mantle range. The Neoarchean to Paleoproterozoic zircon inheritance and core-to-rim increase in δ 18 O observed from some synmagmatic zircons in the Yangpyeong monzonite suggest the interaction  between the lithospheric mantle and basement crust, of which Hf isotopic coupling is indicated by comparable ε Hf (t) values of the zircon cores and rims. As seen in Figure 11, the Gwangcheon monzonite and the Haemi syenite in the western Gyeonggi Massif have distinctly higher zircon ε Hf (t) (-11 to -16) and δ 18 O (8.5-7.6‰) values than those of mafic-intermediate rocks in Yangpyeong and Odaesan, suggesting the fundamental difference of mantle lithosphere between the two areas. Synmagmatic zircons from the granites in the western area (Seosan, Namyang, and Taean) have wide variations in ε Hf (t) (-6 to -21) and δ 18 O (9.8-7.0‰) values, encompassing the ranges of zircon values from the adjacent Gwangcheon monzonite and Haemi syenite. The Hf isotopic range of the source crust could be estimated using the inherited zircon data in the granites. For example, as could be assumed in Figure 10B, the age and Hf isotope data of the Neoproterozoic zircon cores yield moderately negative ε Hf (t = 230 Ma) values (-6.3 ± 2.2) for the Neoproterozoic component in the source crust if the average crustal 176 Lu/ 177 Hf ratio (0.0116; Rudnick and Gao, 2003) is employed. When the typical 176 Lu/ 177 Hf ratio of the mafic crust (0.024) is adopted, the ε Hf (t = 230 Ma) of the source crust increases to -1.7 ± 2.5. These values are comparable with the highest ε Hf (t) of synmagmatic zircon in the Seosan granite (-5.8). Much lower ε Hf (t = 230 Ma) values are calculated for the Mesoproterozoic (-8 to -17 or -1 to -8, depending on the assumed Lu/Hf ratio) and Paleoproterozoic (-32 to -38 or -16 to -21) components. This calculation shows that the Hf isotopic variation of synmagmatic zircons in the Seosan and Namyang granites is attributable to the contribution from diverse age components in the source crust. The correlation between ε Hf (t) and Lu/Hf observed in synmagmatic zircons from the Taean and Namyang granites suggests that the mixing of the crustal components was accompanied with magmatic differentiation (i.e., DePaolo, 1981). The oxygen isotopic evolution of the source crust is more difficult to trace, because lowand high-temperature surface processes could modify the original value significantly (Valley, 2003;Bindeman, 2008;Jo et al., 2016).
On the other hand, oxygen and hafnium isotope data of synmagmatic zircons from the older magnesian group in the Yeongnam Massif reflect the contribution of diverse mantle and crustal reservoirs under arc setting. The lithospheric mantle component could be best represented by zircons from the Andong pyroxenites that formed magmatically in a suprasubduction zone (Whattam et al., 2011). Their δ 18 O values (~7.1‰) are in the range of zircon values from the central-eastern Gyeonggi Massif. However, the near chondritic ε Hf (t) values of the zircons indicate that the chemical fractionation in the lithospheric mantle source beneath the Yeongnam Massif occurred much more recently than in the case of the interior Gyeonggi Massif, possibly during the Paleozoic in association with continued subduction as suggested by the zircon data presented by Kim et al. (2012Kim et al. ( , 2013aKim et al. ( , 2014 and Cheong et al. (2015b). Zircon data for the Yeongdeok adakite represent the isotopic composition of arc crust derived from slab melting. As most adakite samples show (Bindeman et al., 2005), the Yeongdeok zircons have oxygen isotopic compositions close to the normal mantle range. Their ε Hf (t) values are plotted around the evolutionary path of the "arc mantle" projecting from the presentday Hf isotopic composition of island arc rocks (Dhuime et al., 2011) (Fig. 10B). The magmatic reworking of this arc mantle-like Yeongdeok adakite was indicated by the time-ε Hf trend of the zircons from the Late Cretaceous granitoids composing the platform of the Gyeongsang arc system (Cheong and Jo, 2017). The contribution of the Paleoproterozoic crust is evidenced by the presence of coeval inherited zircon cores in the Hamyang granodiorite and the Sangju granite. The negative array of zircon ε Hf (t) and δ 18 O values from the older magnesian plutons (Fig. 11) could be explained by the mixing of these three mantle and crustal components: the metasomatized lithospheric mantle as represented by the Andong pyroxenite (zircon δ 18 O = 7.1 ± 0.3‰, ε Hf = 6.2 to -0.6), young arc crust as represented by the Yeongdeok adakite (zircon δ 18 O = 5.7 ± 0.3‰, ε Hf > 10), and the Paleoproterozoic basement crust (δ 18 O > 7.7‰, ε Hf < -10) contributing most intensively to the Hamyang granodiorite and the Sangju granite. The addition of Precambrian basement rocks into the magma source of the Hamyang granodiorite was probably accompanied with magmatic differentiation, considering the correlation of synmagmatic zircons between ε Hf (t) and Lu/Hf. Zircons from the younger ferroan plutons in the Yeongnam Massif have their own distinct O and Hf isotopic compositions. The low δ 18 O (~5.1‰) and highly positive ε Hf (t) (~8) of zircons from the Hapcheon syenite, higher range of δ 18 O (8.4-6.5‰) and lower but consistently positive ε Hf (t) (~3) of zircons from the Sancheong syenite, and intermediate δ 18 O (~6‰) and slightly positive or negative ε Hf (t) (1.2 to -5.2) of zircons from the Daegang alkali granite may reflect the dominance of asthenospheric mantle, lithospheric mantle, and lower crustal sources, respectively.

Triassic Tectonomagmatic Evolution
The geochronological, geochemical, and O-Hf isotopic data described above allow us to propose a tectonomagmatic model depicted schematically in Figure 12. This model starts with the generation of adakite and sodic magmas along the Yeongnam Massif at ca. 250 Ma.
The Yeongdeok adakite pluton is believed to have been produced under a hot subduction regime possibly triggered by ridge subduction (Yi et al., 2012a) (Fig. 12A). The subduction of the ridge and consequent development of a slab window may have facilitated partial melting of lower crustal rocks that produced the protolith magma of sodic metagranitoids in the adjacent Andong-Cheongsong area. Their REE pattern and zircon ε Hf (t) range (-0.3 ± 2.4) support this interpretation . The subduction of the young and hot ridge likely increased the coupling between the subducting and overriding plates (i.e., Murphy et al., 1998). Under this advancing arc system, magmatism may have ceased or migrated inboard. Based on our new SHRIMP zircon results, after the emplacement of the Yeongdeok adakite at ca. 250 Ma, arc magmatism was temporally shut off in the Yeongnam Massif until the intrusion of the Macheon monzonite at 232 Ma. Interestingly, this reinitiation of magmatism was synchronous with the commencement of potassic and carbonatite magmatism (ca. 233 Ma) in and around the Gyeonggi Massif (Haemi, Namyang, Yangpyeong,Hongcheon,Yangyang,Odaesan,and Goesan;see Fig. 2 for locations) ( Table 2). This correspondence may suggest that the Late Triassic magmatism along the Yeongnam Massif was related to the post-collisional event on the other side.
The Late Triassic plutons in and around the Gyeonggi Massif intruded in a relatively short time interval (<10 Ma). Mafic to intermediate silicate plutons among them are considerably enriched in K and Ba, and are predominantly metaluminous, alkalic to alkali-calcic, and magnesian (Figs. 6 and 7). This rock type, classified as "Caledonian type" (Pitcher 1983), "post-orogenic granitoids" (Maniar and Piccoli 1989), "shoshonitic granitoids" (Duchesne et al., 1998), "K-feldspar porphyritic calc-alkaline granitoids" (Barbarin, 1999), or "appinite" suite (Murphy, 2013), is commonly associated with the post-collisional relaxation (Bonin et al., 1998). As shown in Figure 7A, they invariably display LREE-enriched normalized patterns without prominent Eu anomalies, indicating the presence of garnet or amphibole, but the absence of plagioclase feldspar in the source resi- due. The coeval occurrence of carbonatite and silica-undersaturated potassic and ultrapotassic rocks (Peng et al., 2008;Yang et al., 2010;Kim et al., 2016) manifests that mantle materials were intensively involved in the post-collisional magma. Detachment models to explain the asthenospheric upwelling and its impingement on the lithospheric mantle surface include the convective thinning and removal of the lithospheric mantle (Houseman et al., 1981;Turner et al., 1992) and slab breakoff (Davies and von Blanckenburg, 1995). Of the two models, the slab breakoff may better explain the limited range in time and space for the occurrence of post-collisional plutons in central Korea, as previous works have suggested (Oh et al., 2006b;Seo et al., 2010). In this model, the breakoff of subducted oceanic slab results in upwelling of the hot asthenosphere through the breakage and consequent melting of the metasomatized part of the overriding lithospheric mantle which has relatively low solidus temperatures. As illustrated in Figure 12B, it is suggested that the slab breakoff not only triggered the generation of potassic/ultrapotassic magmas in and around the Gyeonggi Massif but also facilitated the reinitiation of magmatism in the Yeongnam Massif. It is notable that the Macheon monzonite is uniquely shoshonitic in the magnesian Yeongnam plutons (Fig. 6C). Calc-alkaline arc magma then intruded further inland in parts of the Yeongnam Massif and the Okcheon Belt at ca. 225 Ma (Sangju monzonite-granite and Cheongsan-Baegnok granodiorite; see Fig. 2 for locations). The geochemical composition of the 222 Ma Andong ultramafic rocks is consistent with their formation in a suprasubduction zone (Whattam et al., 2011), possibly in response to the sinking of the old, cold, and dense oceanic lithosphere, which left a gap fed by melts flowing upwards from the asthenosphere shortly after the initiation of subduction (Stern and Bloomer, 1992). If this hinge rollback model is correct, the occurrence of the Andong ultramafic complex is indicative of the initiation of a new extensiondominated arc system. The narrow age range (220-217 Ma) and consistent geochemical peculiarities of the ferroan granitoids in the Yeongnam Massif and Okcheon Belt (Ian-Daegang alkali granites and Sancheong-Hapcheon syenites) suggest their common tectonic setting. Ferroan granitoids reflect a close affinity to dry, reduced magmas that mainly, albeit not exclusively, form in extensional environments (Frost et al., 2001;Bonin, 2007). The arc root that had evolved during the continued magmatism at >250-224 Ma likely became negatively buoyant as the mafic-ultramafic cumulate piles at the base of the crust attained mineral equilibration under eclogite facies. We postulate that the formation of the ferroan granitoids resulted from the eventual foundering of such an eclogitized crustal base at 220-217 Ma (Fig. 12C). Considering the limited occurrence of the ferroan granitoids, the size of the drips must have been quite small, which is the case in many areas suspected to have undergone recent delamination (Ducea, 2011, and references therein). This presumed lithospheric delamination and consequent intrusion of the Daegang and Ian A-type granites were virtually coeval with the onset of an extensional arc system indicated by the occurrence of the Andong ultramafic complex, suggesting a close link between the two tectonic events. It is also notable that the emplacement of the Andong complex was only slightly younger than the development of the Gyeonggi Shear Zone (ca. 225 Ma; Kim et al., 2000) that may have resulted from a gravitational collapse of thickened crust. This temporal correspondence indicates that the Gyeonggi and Yeongnam massifs shared an extensional environment in the late stage of Triassic tectonomagmatic evolution.

CONCLUSIONS
This geochronological, geochemical, and O-Hf isotopic study of the Triassic plutons in central and southern Korea reached the following conclusions.
(1) The emplacement of ca. 265-250 Ma calc-alkaline granitoids in southeastern Korea is coeval with or slightly younger than the Barrovian metamorphism (ca. 285-250 Ma) recorded in the Neoproterozoic-Paleozoic fold-and-thrust belts in central Korea. This temporal correspondence suggests a close link between the collisional orogenesis and subduction initiation.
(2) The Late Triassic mangerite-monzonitesyenite-granodiorite-granite plutons emplaced in the Yeongnam Massif could be divided into two geochemically distinct age groups: the older (232-224 Ma) magnesian and alkalicalcic to calc-alkalic group and the younger (220-217 Ma) ferroan and alkalic to alkalicalcic group, which were temporally intervened by a geochemically arc-like ultramafic complex (222 Ma). Zircon O-Hf isotope data indicate that the former was generated by the mixing of the relatively young (most likely Paleozoic) metasomatized lithospheric mantle and arc crust and Paleoproterozoic crust, whereas the latter formed through the melting of asthenospheric/ lithospheric mantle and lower crust.
(3) The inherited zircon age patterns indicate that the infracrustal basements of the interior Gyeonggi and Yeongnam massifs, and the western Gyeonggi Massif are composed of typi-cally North China-like and allochthonous South China-like terranes, respectively.
(4) The Late Triassic K-and Ba-rich, metaluminous, alkalic to alkali-calcic, and magnesian plutons in the Gyeonggi Massif were derived from lithospheric mantle and basement crust. Zircon O-Hf data indicate ancient (most probably Paleoproterozoic) chemical fractionation of the lithospheric mantle beneath the interior Gyeonggi Massif. The fundamental difference of lithospheric mantle between the western and central-eastern Gyeonggi Massif is evidenced by contrasting Hf isotopic compositions of synmagmatic zircons from the mafic-intermediate plutons. The Late Triassic granites in the western Gyeonggi Massif were derived from the Yangtze-like basement crust composed of diverse age components with significantly different Hf isotopic compositions.
(5) Geochronological, geochemical, and O-Hf isotopic data collectively indicate that during the Triassic, the Korean Peninsula experienced complex tectonomagmatic events consisting of: (1) ridge subduction and generation of high-silica adakite and sodic granitoids along the margin of the Yeongnam Massif (ca. 250 Ma), (2) development of an advancing arc system and magmatic quiescence, (3) postcollisional slab breakoff and consequent potassic/ultrapotassic magmatism in and around the Gyeonggi Massif and shoshonitic magmatism in the Yeongnam Massif (ca. 230 Ma), (4) inland migration of arc magmatism until 224 Ma, and (5) tectonic switch to the extension-dominated arc system and delamination of an overthickened arc lithosphere that resulted in the formation of the ferroan plutons in the Yeongnam Massif (222-217 Ma).

ACKNOWLEDGMENTS
We are grateful for the company of Yong-Sun Song and Kye-Hun Park during the sample collection in the Yeongnam Massif. Sook Ju Kim and Guo-Qiang Tang helped in the laboratory work. Insightful reviews by Moonsup Cho and Jacob Mulder improved the manuscript substantially. We also appreciate the careful editorial handling from Rob Strachan. This work was supported by a Korea Basic Science Institute grant (C38709) and a National Research Foundation of Korea grant funded by the government of Korea's Ministry of Science and ICT (2016R1A2B4007283), awarded to Albert Chang-sik Cheong.

SHRIMP Zircon U-Pb Dating
The primary O 2 beam was focused into a ~25-µmdiameter spot at an accelerating voltage of 10 kV. The collector slit width was fixed at 100 µm, achieving a mass resolution of ~5000 at 1% peak height. FC1 (1099 Ma; Paces and Miller, 1993) and SL13 (U = 238 ppm) standard zircons were used for Pb/U calibration and to determine U abundance, respectively. Pb/U ratios were calibrated against FC1 according to the power law relationship between Pb + /U + and UO + / U + . Th/U ratios were estimated using a fractionation factor derived from the measured 232 Th 16 O + / 238 U 16 O + versus 208 Pb/ 206 Pb of the SL13 standard. The common Pb was removed by the 207 Pb (for spots <1000 Ma) or 204 Pb (for spots >1000 Ma) correction method (Williams, 1998) using the model of Stacey and Kramers (1975). Data processing was conducted using the SQUID 2.50 and Isoplot 3.75 programs (Ludwig, 2008(Ludwig, , 2009. Weighted mean ages were calculated after excluding the outliers with the t-test and reported at the 95% confidence level. Spots with high U concentrations (>2000 ppm) were not considered in the age calculation (see White and Ireland, 2012).

Zircon Oxygen Isotope Analysis
The Gaussian focused Cs + primary ion beam was accelerated at 10 kV, with an intensity of ~1.6 nA. The spot size was ~20 µm in diameter (10 µm beam + 10 µm raster). A normal incidence electron flood gun was used to compensate for sample charging. The magnetic field was stabilized using a nuclear magnetic resonance controller. Negative secondary ions were extracted with a -10 kV potential. The field aperture was 6000 × 6000 µm 2 . A ~120 µm entrance slit, 40 eV energy slit, ~133 transfer magnification, and 500 µm exit slit provided a mass resolution of ~2500 at 1% peak height. Under these conditions, the count rate of 16 Owas typically ~1 × 10 9 cps/nA. The 16 O and 18 O ions were detected simultaneously by two Faraday cups with 10 10 and 10 11 Ω resistors, respectively. Measured 18 O/ 16 O ratios were normalized to Vienna standard mean oceanic water ( 18 O/ 16 O = 0.0020052, Baertschi, 1976) and presented as δ 18 O notation. The in-run precision was typically better than 0.2‰ (2 standard errors). The instrumental mass fractionation (IMF) was corrected based on the 91500 zircon value suggested by Valley (2003) (δ 18 O = 10.07 ± 0.03‰). The spot-to-spot reproducibility of 91500 zircon for a single analytical session ranged from 0.3‰ to 0.2‰ (1SD). The FC1 and Penglai zircons yielded average IMF-corrected δ 18 O values of 5.89 ± 0.21‰ (n = 42, 1SD) and 5.28 ± 0.26‰ (n = 47, 1SD), respectively. The running results for the Penglai zircon standard (see Table DR4) were consistent with the recommended value (5.31 ± 0.10‰; Li et al., 2010).

Zircon Lu-Yb-Hf Isotope Analysis
Ten Faraday collectors were set to simultaneously detect the required isotopes: 172 Yb, 173 Yb, 174 (Hf+Yb), 175 Lu, 176 (Hf+Lu+Yb), 177 Hf, 178 Hf, 179 Hf, 180 Hf, and 182 W. Data were acquired by ablating 50-µm-diameter laser spots. The instrument parameters were a 10 Hz repetition rate and an energy density of 6-8 J/cm 2 . Helium (650 mL/min) and N 2 (2 mL/min) were used as carrier gases to obtain high Hf isotope intensities (Iizuka and Hirata, 2005). The interferences of 176 Lu and 176 Yb on the 176 Hf signal were corrected using isotopic values suggested by Chu et al. (2002) and Vervoort et al. (2004), respectively. The mass bias of the measured Hf isotopic ratios was corrected to 179 Hf/ 177 Hf = 0.7325 using an exponential correction law. The 176 Lu/ 177 Hf and 176 Yb/ 177 Hf ratios were calculated following the method of Iizuka and Hirata (2005). The initial ε Hf values were calculated using a 176 Lu decay constant of 1.865 × 10 -11 year -1 (Scherer et al., 2001) and the chondritic values suggested by Blichert-Toft and Albarède (1997). The raw data were processed using Iolite 2.5 running within Igor Pro 6.3.5.5 software (Paton et al., 2011), and corrected for the background, laser-induced elemental fractionation, and mass discrimination. All ratios were calculated with 2σ errors. During the sample analysis, 91500 and FC1 zircons were repeatedly analyzed at the beginning and end of each analytical session, and at regular intervals during each session. The running results of zircon standards (91500 and FC1, see Table DR4) were consistent with the recommended values in the literature (Griffin et al., 2000;Kemp et al., 2010) within uncertainties.