The waning stage of a long-lived collisional orogeny is commonly governed by an extensional regime in association with high-temperature metamorphism, anatexis, and magmatism. Such a late-orogenic process is well-recorded in the Okbang amphibolite, Yeongnam Massif, Korea, where thin layers or irregular patches of tonalitic leucosomes are widespread particularly in association with ductile shear zones. Various microstructures including interstitial felsic phases and former melt patches indicate that leucosomes are the product of partial melting. These leucosomes are aligned en echelon and contain large (up to ~2 cm) grains of peritectic hornblende, suggesting synkinematic fluid-present anatexis. The leucosomes are enriched in Na2O and Sr contents compared to the amphibolite but depleted in rare earth and high field-strength elements. P-T conditions of the anatexis were estimated at 4.6–5.2 kbar and 650–730°C, respectively, based on hornblende-plagioclase geothermobarometry. Sensitive high-resolution ion microprobe U-Pb analyses of zircon from an amphibolite and a leucosome sample yielded weighted mean 207Pb/206Pb ages of 1866±4Ma and 1862±2Ma, which are interpreted as the times for magmatic crystallization and subsequent anatexis of mafic protolith, respectively. The latter is consistent with the time of partial melting determined from a migmatitic gneiss and a biotite-sillimanite gneiss at 1861±4Ma and 1860±9Ma, respectively. The leucosomes are transected by an undeformed pegmatitic dyke dated at 1852±3Ma; by this time, extensional ductile shearing has ceased. Initial εHft values of zircon from the amphibolite range from 4.2 to 6.0, suggesting juvenile derivation of basaltic melt from the mantle. In contrast, lower εHft values (–0.1 to 3.5) in leucosome zircons indicate a mixing of crust-derived melt. Taken together, the Okbang amphibolite has experienced synkinematic fluid-present melting during the waning stage of Paleoproterozoic hot orogenesis prevalent in the Yeongnam Massif as well as the North China Craton.

Partial melting of amphibolites at middle-lower crustal depths commonly takes place in response to the dehydration of amphibole (e.g., [15]) or the influx of externally derived hydrous fluid (e.g., [69]). This anatexis accounts for the widespread occurrence of tonalitic and trondhjemitic melts in orogenic belts [1015], and the migration of these melts provides the mechanism for crustal reworking to yield residual granulites or restites in the continental crust (e.g., [16, 17]). Thus, melt formation and migration are two principal processes responsible for chemical differentiation of the continental crust as well as substantial variation in the crustal strength (e.g., [1821]). Many previous field-based studies have focused on the dehydration melting of amphibolites [2, 5, 22] but much less on the fluid-present melting, although the latter is one of the key melt-forming processes in the middle to lower crust during orogenesis [1, 6, 7, 9, 23, 24]. Nevertheless, fluid-present melting, resulting in a volume decrease in host rocks, is associated with plate margin tectonic structures such as crustal-scale shear zones or regional thrusting that may lead to extensional collapse and exhumation of an orogenic belt [2527]. Thus, both fluid-present and fluid-absent melting processes are important for understanding the melt formation, migration to drainage, and crustal rheology during orogenesis [27].

The Okbang amphibolite in the Yeongnam Massif, Korea (Figure 1) contains abundant tonalitic leucosomes with a variety of microstructures and provides a natural laboratory for investigating the processes involved in fluid-present melting and melt migration. Based on the whole-rock geochemistry and multigrain zircon U-Pb age, previous workers [28, 29] suggested that the protolith of amphibolites formed in a rift-related setting at ~1.92 Ga, but these authors failed to recognize partial melting in the amphibolite. In order to assess the role of fluid-present melting, we investigated a suite of amphibolites, neosomes, and host gneisses (for the general terminology of migmatites; we followed the recommendation of Sawyer [8]) to determine field relationships, bulk-rock geochemistry, and zircon U-Pb ages using a sensitive high-resolution ion microprobe (SHRIMP). In addition, the Lu-Hf isotopic compositions of zircon were analyzed to unravel the crust-mantle interaction during the formation of felsic melts in the amphibolite. Our results, combined with available data, provide further insight into ~1.87–1.85 Ga tectonomagmatism in the Korean Peninsula and its linkage to the prolonged Paleoproterozoic orogenesis of the North China Craton [3033].

The Korean Peninsula consists of three major Precambrian massifs (Nangrim, Gyeonggi, and Yeongnam) adjoined by the Gyeonggi Marginal Belt ([30], Figure 1(a)). The Yeongnam Massif is a polymetamorphic terrain bounded to the north by the Ogcheon Belt and unconformably overlain to the southeast by thick volcanic-sedimentary sequences of the Cretaceous Gyeongsang Basin. This massif is primarily composed of quartzofeldspathic gneiss, migmatitic gneiss, porphyroblastic gneiss, and granitic gneiss together with lesser amounts of amphibolite and calc-silicate rocks [34]. Igneous protoliths of gneisses were largely emplaced at ~2.0–1.9 Ga in an arc-related environment and subsequently affected by upper amphibolite to lower granulite facies metamorphism at ~1.9–1.85 Ga (Figure 1(b); [30, 31, 33, 35, 36]). The latter event is associated with widespread partial melting to produce abundant leucosomes and garnet-bearing leucogranites and perhaps best exemplified by the granulite-facies aureole around 1.87–1.86 Ga anorthosite-mangerite-charnockite-granite (AMCG) suite in the southern Yeongnan Massif (Figure 1(a); [32, 33]). P-T conditions of migmatitic gneisses in the Yeongnam Massif are estimated at 4–6 kbar and 750–850°C, based on conventional geothermobarometry and phase equilibria modeling [30, 33, 34]. Similar high-temperature tectonothermal events have been well-documented in the North China Craton at 2.0–1.8 Ga, including the Trans-North China Orogen where high thermal gradients were attained during accretionary–collisional orogenesis [3739].

The Okbang amphibolite, the focus of this study, occurs as planar bodies along the boundary between the Buncheon granitic gneiss and metasedimentary rocks (Figure 2; [40]). The Buncheon gneiss typically contains augens of megacrystic K-feldspar and is penetratively deformed on an outcrop scale. On the other hand, the metasedimentary rocks are mainly composed of interbedded pelitic and psammopelitic gneisses together with a lesser amount of quartzite. In particular, the metapelitic rocks define high dT/dP, Buchan metamorphic zones progressing from cordierite through sillimanite to garnet zones; the latter two zones are associated with widespread anatexis typified by the occurrence of abundant (up to ~20 vol.% on the outcrop) cordierite-bearing leucosomes and leucogranites [34, 35, 41]. The country rocks hosting the Okbang amphibolite are strongly deformed, particularly in the vicinity of lithologic boundaries (Figures 3(a) and 3(b)). Extensional (C) shear bands typically occur as sets of closely spaced ductile shear zones at millimeter to centimeter scales, accompanied by partial melting and leucosome formation. Synkinematic melting and melt segregation may be accounted for in the context of an extensional tectonic regime prevalent during the Paleoproterozoic in the southern Yeongnam Massif [31, 33]. Whole-rock geochemical analyses revealed the rift-related, enriched mid-ocean ridge basalt (E-MORB) composition of amphibolite [28], and the U-Pb zircon dating based on multigrain thermal-ionization mass spectrometry analysis yielded a discordant date of 1918±10Ma [29]. In addition, precise SHRIMP U-Pb zircon age constraints are available for the Buncheon granitic gneiss, i.e., magmatic crystallization at 1966±16Ma and subsequent metamorphism at 1862±4Ma [35, 42]. The latter is associated with late orogenic event which is widespread in the entire Korean Peninsula, including the Yeongnam Massif [30, 36].

The Okbang amphibolite occurs as elongate lenses ranging in widths from ~20 to 400 m that are stretched and folded on a map scale (Figure 2). Various lines of evidence for partial melting are present on the outcrops, and metatexitic amphibolites are predominant (Figures 3 and 4; [8]). Penetrative foliations (S1) in the amphibolite are occasionally defined by preferred orientation of hornblende neoblasts and layer-parallel leucosomes; their attitudes are consistent with those measured in the Buncheon and metasedimentary gneisses generally striking northeast and dipping moderately to the northwest (Figure 2). Both stretching and mineral (hornblende) lineations mostly plunge to the north. Towards the boundary with the Buncheon granitic gneiss, the amphibolites are progressively deformed to yield high-strain zones where leucosomes are highly stretched and isoclinally folded. The metasedimentary rocks also contain abundant granitic leucosomes and garnet–tourmaline-bearing leucogranites, but the Buncheon granitic gneiss is lacking in melt-related features probably owing to the limited amount of fluid.

Tonalitic leucosomes in the amphibolite are generally aligned parallel to the foliation but locally occur as discordant or patchy segregations (Figure 3(c)). Some layer-parallel leucosomes are deflected into the interboudin partition to yield the “collapse” structure [43]. The proportion of leucosomes on many amphibolite outcrops is generally in the range of ~20–30 vol.% but rarely reaches up to ~40 vol.%. Moreover, these estimates significantly vary on individual outcrops, ranging from <5 to ~30 vol.% over a distance of a few tens of meters. In situ leucosomes have diffuse boundaries with volumetrically minor melanosomes enriched in hornblende (Figure 3(d)). The leucosomes, either concordant or discordant, are interconnected with each other, and often concentrated along the axial plane of mesoscopic folds and within the interboudin partition of amphibolite, suggesting synkinematic formation of melts (Figures 3(c), 3(e), 3(f), 4(a), and 4(b); [44]). On the other hand, poikilitic megacrysts of hornblende (up to 2 cm) are common in the leucosomes where melanosomes are generally lacking (Figure 4(c)). We interpret that these megacrysts represent peritectic hornblende-rich segregations interconnected with tonalitic leucosome pools or layers (Figures 4(d) and 4(e)).

The cessation of synkinematic melting was followed by the intrusion of discordant pegmatitic dykes and pods. The pegmatitic bodies range from a few centimeters to ~3 m in width, lack penetrative deformation, and often transect the stringers of leucosomes aligned subparallel to the foliation in amphibolite (Figure 4(f)). The margins of pegmatites often contain alteration bands. In particular, tungsten ore deposits are associated with pegmatites and attributed to metasomatic reactions of the amphibolite with tungsten-bearing alkaline melts [40, 45]. The Okbang ores containing ~70% WO3 consist of scheelite, wolframite, fluorite, and minor sulfide minerals and represent a rare example of pegmatite-type tungsten ore worldwide [46].

The Okbang amphibolites are medium to coarse-grained and primarily consist of hornblende, plagioclase, biotite, and quartz together with small amounts of ilmenite and magnetite. Zircon and apatite occur as accessory phases. Microstructures with straight grain boundaries and ~120° triple junctions are characteristic (Figure 5(a)), and prismatic grains of hornblende are often aligned parallel to a weak foliation. The hornblende grains are mostly subhedral, rarely exceed 1.5 mm in the maximum dimension, and commonly contain rounded inclusions of quartz and rare plagioclase. Plagioclase is generally stubby, subhedral to anhedral, and less than 1.0 mm in size (Figure 5(a)). The majority of plagioclase grains are compositionally zoned and show undulose extinctions and/or deformation twins. Quartz grains are also subhedral to anhedral and less than 0.5 mm in size and show undulose extinction. In particular, interstitial quartz fills the space among rounded grains of hornblende and plagioclase (Figures 5(b) and 5(c)); this microstructure is indicative of dissolution at high temperatures [8]. Biotite uncommonly occurs as subhedral, medium-grained (up to 5 mm in size) poikiloblasts, containing small inclusions of hornblende and plagioclase together with interstitial quartz (Figure 5(d)). Thin films of quartz develop to mantle the rounded grains of hornblende and plagioclase in contact with biotite (Figure 5(e)); such microstructures are diagnostic of the former presence of in situ melts [47].

The leucosomes in amphibolites primarily consist of plagioclase and quartz; both phases commonly reach ~90 vol.%. Mafic phases are represented by poikilitic hornblende containing inclusions of plagioclase, quartz, and rare biotite (Figure 5(f)). Plagioclase grains are mostly subhedral and locally form framework structures with interstitial quartz aggregates (Figure 5(g)); such a microstructure is interpreted to denote the presence of former melts crystallized into the leucosome [48]. In addition, equant plagioclase grains together with irregular morphology of quartz suggest minor influence of postsolidification strain [8]. Hornblende-rich segregations consist of megacrystic hornblende (up to ~1.5 cm in the maximum dimension) together with medium-grained plagioclase and quartz.

The melanosomes are generally enriched in hornblende together with minor plagioclase and quartz. Hornblende is subhedral to anhedral and commonly contains quartz inclusions; its grain size ranges from a few millimeters to ~2 cm, significantly greater than that of the amphibolite (Figure 5(h)). Interstitial or vermicular quartz uncommonly occurs along the grain boundaries of hornblende; such microstructures are interpreted to represent the pseudomorph after former melts or the reaction product between hornblende and residual melt [8, 49].

5.1. Analytical Methods

Major element compositions of calcic amphibole and plagioclase were determined using a JEOL JXA-8100 electron microprobe at Gyeongsang National University, Korea, with an accelerating voltage of 15 kV and a beam current of 10 nA. The beam diameter was typically 5 μm, but a wider beam of 10 μm was used for analyzing feldspars to minimize the loss of Na. Analytical errors are generally less than 2%. Data acquisition and reduction were performed using the ZAF (Z, atomic number; A, X-ray absorption; and F, secondary fluorescence effects) calculation for the matrix correction. Representative analyses for amphibole and plagioclase are given in Tables 1 and 2.

Representative samples of amphibolite, leucosome, melanosome, and hornblende segregation were analyzed for major and trace elements in order to determine geochemical variations during the melting. Samples were crushed to millimeter-sized fragments and then ground in a tungsten carbide ring mill to powders less than 125 μm in size. Major elements were determined on the fused lithium tetraborate (Li2B4O7) glass beads using a Panalytical Axios Advanced wavelength dispersive X-ray fluorescence spectrometry (XRF) at Activation Laboratories (Actlabs), Canada. Trace and rare-earth elements were measured using a PerkinElmer Sciex ELAN 9000 inductively coupled plasma-mass spectrometry (ICP-MS) at the Actlabs. Analytical uncertainties are in the range of 1–3%. Representative compositions of amphibolite, leucosome, melanosome, and hornblende segregation are listed in Table 3.

5.2. Mineral Chemistries and P-T Conditions

5.2.1. Calcic Amphibole

The analyzed amphiboles are mostly magnesiohornblende with an Si value of 6.93–7.27 a.p.f.u. (atoms per formula unit) and XMg=Mg/Mg+Fetotal of 0.51–0.60, according to the nomenclature of Leake et al. [50]; those from the segregation domain (Figure 4(e)) lie in the field of ferrohornblende with lower Si (6.53–6.66 a.p.f.u.) and XMg (0.39–0.42; Figure DR1). These hornblendes are mostly homogenous, although the hornblende cores in the amphibolite could be slightly lower in XMg (0.51–0.60) but higher in Ti content (0.07–0.13 a.p.f.u.) than those of the rims (0.55–0.60 and 0.04–0.11 a.p.f.u., respectively).

5.2.2. Plagioclase

The analyzed compositions of plagioclase show distinctive compositional ranges (Figure DR2); the anorthite contents in amphibolite are in the range of An47–57, intermediate between those of melanosome (An52–59) and leucosome (An43–47). On the other hand, the orthoclase contents are highly variable and range from Or4 to Or25. In the amphibolite, plagioclase grains are commonly zoned, with the An content decreasing towards the rim (Figure DR2). The rim compositions are compatible with those of the leucosome, probably reflecting the reequilibration during the cooling. Plagioclase grains of the leucosome are mostly homogeneous and contain less An than those of the melanosome (Figure DR2). Such a relationship is consistent with the anatectic origin of leucosomes [51, 52]. It is further noted that An content (An30–55) of plagioclase inclusions within large biotite poikiloblasts is the most variable among the analyzed grains.

5.2.3. P-T Conditions

The Okbang amphibolite has garnet-free assemblages of high variance, hampering any attempt for precise determination of P-T conditions. Nevertheless, we estimated the P-T conditions using the hornblende–plagioclase geothermometer [53, 54] in combination with the hornblende–plagioclase–quartz geobarometer governed by a net-transfer reaction, tremolite+tschermakite+2albite=2pargasite+8quartz [55]. The result using the compositions of hornblende-plagioclase cores gives the P-T range of 4.6–5.2 kbar and 650–730°C (Figure DR3), which is consistent with P-T conditions of previous studies determined from cordierite-bearing metapelitic assemblages in the northern Yeongnam Massif [34, 41]. Our P-T estimates are also consistent with those of wet solidus in mafic rocks based on the pseudosection modeling [56] or experimental calibrations [11, 14, 57, 58]. In contrast, dehydration melting of amphibole occurs at temperatures higher than ~850°C, yielding anhydrous peritectic phases such as orthopyroxene or clinopyroxene [13, 14, 56, 59]. Such anhydrous minerals are absent in the Okbang amphibolite, whereas hornblende ubiquitously occurs in the leucosome. Hence, we conclude that tonalitic leucosomes in the amphibolite were produced by fluid-present melting at midcrustal conditions.

5.3. Whole-Rock Geochemistry

5.3.1. Major and Trace Elements

The amphibolites, leucosomes, and melanosomes generally show linear relationships in major element contents (Figure 6). The compositions of three amphibolites together with those of previous studies [28, 60] are in the range of 48–53 wt.% SiO2, 12.2–14.7 wt.% Al2O3, and 6.8–11.8 wt.% CaO; the Mg number =Mg/Mg+Fe and TiO2 content are 0.2–0.5 and 0.6–2.0 wt.%, respectively. Their compositions mostly belong to gabbro and gabbroic diorite in the total alkali vs. silica diagram [61] and are plotted in the tholeiitic field on the AFM diagram [62]. In contrast, leucosomes are rich in silica (>70 wt.% SiO2) and weakly metaluminous (A/CNK=0.800.98); their Na2O (3.0–3.6 wt.%) contents are higher than K2O (0.2–0.5 wt.%). Relative to amphibolites, leucosomes are enriched in SiO2, Al2O3, and Na2O contents but depleted in TiO2, Fe2O3, MgO, and CaO (Table 3). Major element concentrations of leucosomes are similar to those of melts derived from wet melting of amphibolites and dioritic rocks (Figure 6; [15, 23]). Melanosomes have similar TiO2, CaO, and K2O but higher Fe2O3+MgO abundances compared to amphibolites (Figure 6).

The compositions of leucosomes fall into the tonalite field on normative An-Ab-Or diagram (Figure 7) and are compatible with those of partial melts experimentally produced by fluid-present melting of tholeiitic basalt [63]. Our leucosome compositions are also comparable to those of melt patches derived from the fluid-present melting of amphibolites in the Gyeonggi Massif and western Shandong Province, China (6; [24]). In contrast, a series of phase equilibria modeling with a minimal amount of fluid suggests that the partial melting of mafic rocks yields granitic melts at relatively low temperatures (Figure 7; [13]). Thus, fluid-deficient open-system melting cannot account for tonalitic melts of the Okbang amphibolite.

On the primitive mantle-normalized spider diagram, the amphibolites are characterized by negative Nb-Ta anomalies, together with weak Zr, Hf, and Ti anomalies (Figure 8(a)). Hornblende-free leucosomes are enriched in Sr compared to the amphibolite, but depleted in Rb; in contrast, hornblende-bearing leucosomes are enriched in Rb. High Sr/Rb in the former indicates the prevalence of feldspar over mica participating in the melting reaction [64]. High field-strength elements (HFSE) such as Zr, Hf, and Ti are generally enriched in melanosomes and hornblende-rich segregations compared to amphibolites, suggesting the accumulation of accessory phases in restites during melt extraction. In contrast, the leucosomes have the lowest HFSE concentrations and are depleted in other trace elements such as Y, Sn, V, Cr, and Ni, which are generally partitioned into ferromagnesian minerals (Table 3).

5.3.2. Rare Earth Elements

The analyzed amphibolites yielded a relatively flat REE pattern at ~10–20 times chondrite, with a slight LREE enrichment (Figure 8(b); LaN/SmN=1.161.33) and small negative Eu anomalies Eu/EuN=EuN/SmN/GdN=0.790.98. Such a result is consistent with that of previous studies, suggesting that protolith of the amphibolite was a tholeiitic basalt (Figure 8(b); [28]). The leucosomes are typified by fractionated REE patterns with (La/Yb)N of 2.84–19.60 and positive Eu anomalies (Eu/Eu=1.546.88; Figure 8(b)). The latter, together with the Sr enrichment (Table 3), is compatible with the plagioclase accumulation (Figure 5(g)). In contrast, hornblende segregations are characterized by slightly convex-upward middle REE and pronounced negative Eu anomaly (Eu/Eu=0.29); both of which are diagnostic of hornblende crystals precipitating from tonalitic melt [65, 66]. Hence, the segregation is best interpreted as hornblende-rich restite that has trapped a small amount of melt [67]. Melanosomes, similar to the amphibolite, yielded relatively flat REE patterns and weak negative Eu anomalies (Eu/Eu=0.710.90; Figure 8(b)).

6.1. Analytical Methods

Zircon grains for the U-Pb isotopic analyses were separated from six samples: amphibolite (UJ03), leucosome (OB-01L), pegmatitic dyke (UJ03-3), Buncheon granitic gneiss (BCG-1), migmatitic gneiss (WN-1), and biotite-sillimanite gneiss (WN-2); the latter two are from the metasedimentary rock unit (Figure 2). Conventional heavy liquid and hand picking techniques were used for the separation, and zircon grains were mounted on a 25.4 mm epoxy disk together with FC1 zircon standard (1099 Ma; [68]). The mount was ground and polished to expose the approximate centers of zircon grains. Cathodoluminescence (CL) images of individual grains were obtained using a JEOL JSM 6380 scanning electron microscope at the School of Earth and Environmental Sciences, Seoul National University. In situ U-Pb zircon ages were obtained using a SHRIMP housed at the Korea Basic Science Institute (KBSI). The analytical protocol for zircon follows routine procedures described by Williams [69]. All isotopes were acquired using a negative ion oxygen (O2) beam. The primary oxygen beam was 4–6 nA in intensity and ~25 μm in diameter. The measured Pb/U and Pb/Th ratios were corrected using the reference zircon FC1. The abundances of U, Th, and Pb were normalized to the value (U=238ppm) of standard zircon SL13. The common Pb contributions were corrected using the measured 204Pb amount and the model common Pb composition [70]. The Squid 2 and Isoplot/Ex softwares [71, 72] were used for the age calculation and data evaluation, respectively. Individual spot analyses and weighted mean 207Pb/206Pb ages are quoted at 1σ and 2σ confidence levels, respectively. Analyses with large uncertainty (>10%) and discordance (>5%) were discarded in the age calculation.

After completing the U-Pb analyses, the same zircon amounts were used for in situ Hf analyses employing a Nu Plasma II MC-ICPMS combined with a 193 nm ArF excimer laser ablation system housed at the KBSI. Hf isotopic compositions of zircon were measured on top of the U-Pb analytical pit (Figures 9(a)–9(c)). The ablation time for each analysis was ~60 s, with a 5 Hz repetition rate, and the beam diameter was ~50 μm. Lu and Yb isotopic compositions, employed for the correction of mass bias and isobaric interference, were adopted from Vervoort et al. [73] and Chu et al. [74], respectively. Isobaric interference-corrected 176Hf/177Hf ratios were exponentially normalized to 179Hf/177Hf=0.7325 [75]. Zircon 91500 and FC1 were used as standards with recommended 176Hf/177Hf ratios of 0.282306±8 and 0.282184±16, respectively [76, 77]. The 176Lu/177Hf and 176Yb/177Hf ratios were calculated after Iizuka and Hirata [78] and used for estimating initial 176Hf/177Hf isotopic ratios and corresponding initial epsilon Hf εHf=176Hf/177Hfsample/176Hf/177Hfchondrite1×104 values and Hf model ages. Analytical results were monitored on 176Yb/177Hf and 176Lu/177Hf ratios in order to check isobaric interferences from Yb and Lu, respectively. Initial εHft values and model ages (TDM1) were calculated using a 176Lu decay constant of 1.865×1011y1 [79] together with depleted mantle (176Lu/177Hf=0.0384, 176Hf/177Hf=0.28325) and chondritic (176Lu/177Hf=0.0332, 176Hf/177Hf=0.282772) values of Griffin et al. [80] and Blichert-Toft and Albarède [81], respectively. Two-stage model ages (TDM2) were calculated with reference to the parameters suggested for lower crust (176Lu/177Hf=0.0187; [82]) and depleted mantle (176Lu/177Hf=0.0384; [83]).

6.2. Zircon Morphologies and U-Pb Ages

Zircon grains from the amphibolite (UJ03) are ~50–200 μm in size and stubby in shape with aspect ratios ranging from 1 to 1.5 (Figure 9(a)). Most grains have broad-banded or irregular CL zoning and are devoid of an overgrowth rim. Uranium concentrations and Th/U ratios of zircon range from 172 to 335 ppm and 0.16 to 0.34, respectively, corroborating an igneous origin ([84]; Table DR1). A few zircon grains develop thin dark-CL rims which are apparently homogeneous without pores or inclusions (<30 μm; Figure 9(a)), suggesting that these rims have overgrown primary zircons during the melt crystallization [85]. The U contents of zircon rims are in the range of 205–2431 ppm, and their Th/U ratios vary from 0.07 to 0.23. Twelve spot analyses of magmatic zircons yielded a weighted mean 207Pb/206Pb age of 1866±4Ma (MSWD=0.58; Figure 10(a)), whereas three analyses of dark-CL rims gave 1861±3Ma (MSWD=0.27; Figure 10(a)). The latter is interpreted as the time of zircon precipitation from the melt.

Zircons from the tonalitic leucosome (OB-01L) are ~100–200 μm in size and have subhedral to anhedral with aspect ratios ranging from 1.5 to 2.5 (Figure 9(b)). They have irregular or broad-banded zoning with dark-CL cores and oscillatory-zoned or structureless overgrowth rims; the cores contain high U, reaching up to 4548 ppm (Table DR1). Th/U ratios of the rims are limited to the range of 0.04–0.06 (Table DR1). The weighted mean 207Pb/206Pb age of twelve rims, excluding three analyses possibly affected by Pb loss or inherited core component, is 1862±2Ma (MSWD=0.96; Figure 10(b)), corresponding to the time of melt crystallization.

Zircon grains from undeformed pegmatitic dyke (UJ03-3) are euhedral or subhedral and relatively large (up to ~500 μm; Figure 9(c)). Their aspect ratios range from ~1.5 to 3.0. Most grains show homogeneous CL features with high U contents (2382–8938 ppm), and their Th/U ratios are below 0.01 (Table DR1). These high-U zircons generally yield older than true 206Pb/238U ages owing to the machine-induced bias related to metamictization of grains, but the 207Pb/206Pb ages are unaffected by high-U bias [86]. The weighted mean 207Pb/206Pb age of five grains, excluding two spot analyses affected by possible inheritance or Pb loss, is 1852±5Ma (MSWD=2.5; Figure 10(c)) which is interpreted as the time of crystallization. Many zircon analyses show a slightly reverse discordance which is attributable to the site-specific matrix effect on damaged, metamict, and high-U zircons with unsupported radiogenic Pb [87]. Nevertheless, the 207Pb/206Pb ages estimated from reverse discordant analyses are consistent with concordant ages [88].

Zircon grains from the granitic gneiss (BCG-1) are ~50–200 μm in size and euhedral to anhedral with aspect ratios of 1.0–2.5 (Figure 9(d)). Their CL images are often homogeneous or reveal broad-banded and oscillatory zones. Zircon grains contain mineral inclusions such as quartz and thorite associated with some pores, probably resulting from fluid-induced recrystallization [89]. The analyzed magmatic zircons have variable U contents (967–4673 ppm) and Th/U ratios (0.17–0.41; Table DR1). In contrast, recrystallized domains have low Th (38–127) and moderate U (1102–3024 ppm) contents, yielding low Th/U ratios of 0.04–0.05. The recrystallized domains yielded 207Pb/206Pb dates varying from 1953±4Ma to 1900±5Ma (Figure 10(d)) and were excluded for further consideration. Excluding these six analyses, twelve spot analyses of magmatic zircons yielded a weighted mean 207Pb/206Pb age of 1981±2Ma (MSWD=0.68; Figure 10(d)) which is interpreted as the time of magmatic crystallization.

Zircon grains from the migmatitic gneiss (WN-1) are typically ~70–200 μm in size with aspect ratios ranging from 1.5 to 2.0 (Figure 9(e)). The inherited cores are commonly oscillatory zoned and overgrown by homogeneous or weakly zoned, dark-CL rims; such a microstructure is typical for metamorphic overgrowth (e.g., [30, 90]). The inherited cores yielded 207Pb/206Pb dates varying from 2717±9Ma to 2092±4Ma (Figure 10(e)). Th/U ratios of the cores range from 0.13 to 1.03 except for one spot analysis, whereas those of the rims are smaller than 0.01, suggesting metamorphic origin (e.g., [84, 91]). Eight analyses of the rim yielded a weighted mean 207Pb/206Pb age of 1861±4Ma (MSWD=0.51; Figure 10(e)) consistent with the time of leucosome crystallization.

Zircons from the biotite-sillimanite gneiss (WN-2) are typically ~100–200 μm in size with aspect ratios of 1.5–3.5 (Figure 9(f)). The inherited cores of zircon reveal oscillatory or patchy zoning and are rarely overgrown by homogeneous dark-CL rims. The inherited cores yielded 207Pb/206Pb dates widely varying from 2700±7Ma to 2019±5Ma (Figure 10(f)), suggesting that the depositional age of sedimentary protolith is younger than ~2.02 Ga. Th/U ratios of the cores range from 0.09 to 1.28 except for one analysis, whereas those of the rims are less than 0.01 (Table DR1). The weighted mean 207Pb/206Pb age of two rims, excluding one analysis affected by Pb loss, is 1860±9Ma (MSWD=0.39; Figure 10(f)). These results are consistent with those of migmatitic gneiss.

6.3. Zircon Hf Isotopic Compositions

Hafnium isotopic compositions of the analyzed zircons are listed in Table DR2 and shown in Figure 11. Initial 176Hf/177Hf ratios of zircons from the amphibolite (UJ03) are in the range of 0.281723–0.281787, and their εHft values range from 4.2 to 6.0, when calculated using the weighted mean 207Pb/206Pb age. Depleted-mantle model ages and two-stage Hf model ages are in the range of 2.10–2.03 Ga and 2.22–2.13 Ga, respectively (Table DR2). In contrast, zircons from the leucosome (OB-01 L) have initial 176Hf/177Hf ratios (0.281629–0.281716) and εHft values (–0.1 to 3.5) smaller than those of UJ03. These leucosome zircons have depleted-mantle model ages and two-stage Hf model ages of 2.27–2.14 Ga and 2.50–2.29 Ga, respectively. Initial 176Hf/177Hf ratios of zircon from the pegmatitic dyke (UJ03-3) are in the range of 0.281544–0.281639, corresponding to εHft values of –4.7 to –2.8, and reflect old crustal component inherited from the metasedimentary host (–11.9 to –6.1 εHft; [35]). Depleted-mantle model ages and two-stage Hf model ages of pegmatite are in the range of 2.46–2.41 Ga and 2.75–2.64 Ga, respectively (Table DR2).

7.1. Evidence for Fluid-Present Melting

The H2O-fluxed melting of mafic rocks is a fundamental process in the orogenic belt where the development of crustal-scale shear zones or the underplating of arc-related magmas may lead to the infiltration of aqueous fluid into heated crust [26, 27]. The Okbang amphibolite, a product of mafic magmatism associated with the Paleoproterozoic orogeny in the Yeongnam Massif [28, 29, 31], reveals various lines of evidence for H2O-fluxed melting associated with ductile deformation (Figures 2 and 3). Firstly, the leucosomes are typified by the presence of poikilitic hornblende megacrysts containing bleb-like inclusions of plagioclase, quartz, and rare biotite (Figure 4(c); [23]) in the absence of anhydrous peritectic phases. Such microstructural relationship is accounted for by an H2O-fluxed melting reaction: hornblende1+plagioclase+quartz±biotite+H2O=hornblende2+melt.

This reaction is consistent with the melt-producing reaction of Lappin and Hollister [92] and generally accompanied by a net decrease in volume with negative dP/dT slope [26, 93]. In particular, negative reaction volume may lead to tensional microcracking that could facilitate local melt accumulation (Figure 3(e); [94]). Secondly, the occurrence of quartz interstitial to the rounded grains of hornblende and plagioclase (Figure 5(b)–5(e)) corroborates the presence of excess fluid during the melting because an increase in the H2O content of incipient melt may reduce the wetting angle [95, 96]. Finally, hornblende-rich melanosomes or restites are volumetrically minor and locally present as narrow slivers along the leucosome margin (Figures 3(d) and 3(e)). These melanosomes are attributed to an H2O-fluxed melting because hornblende occurs as a peritectic phase in the absence of other mafic phases (e.g., [2527]). In addition, experimental studies suggest that peritectic amphiboles are stable for melting in the mafic rocks, with a minimum H2O amount of ~4 wt.% at 2 kbar or ~2.5 wt.% at 8 kbar [97, 98]. Gardien et al. [99] also reported that the addition of H2O into the melt has stabilized the amphibole solid-solution. Thus, the predominance of peritectic hornblende in leucosomes is attributed to the presence of external aqueous fluid during the melting. Progressive dehydration of metaturbiditic sequences hosting the Okbang amphibolite might be responsible for the supply of free aqueous fluids necessary for melting [34, 100]. Further studies are needed for a better comprehension of fluid sources for the melting.

7.2. Timing of Mafic Magmatism and Fluid-Present Melting

Multiple episodes of zircon crystallization in amphibolites and associated leucosomes permit us to constrain the times of mafic magmatism and high-temperature metamorphism, respectively. A variety of field relationships, including diffuse boundaries between the leucosome and host amphibolite (Figure 3(c)) and petrographic continuity between boudin necks and layer-parallel leucosomes (Figures 3(e) and 4(a)), suggest that synkinematic melting is associated with the segregation of tonalitic melt. These leucosomes are subsequently transected by undeformed pegmatitic dykes (Figure 4(f)).

The zircon cores and rims from the amphibolite yielded the weighted mean 207Pb/206Pb ages of 1866±4Ma and 1861±3Ma, respectively, which are best interpreted as the times for magmatic emplacement and subsequent metamorphism, respectively. The development of overgrowth rims may be attributed to in situ melting of the amphibolite, as evidenced by the presence of melt-related microstructures (Figures 5(b) and 5(c)). This melting episode is coeval with the anatexis in leucosomes dated at 1862±2Ma. In addition, the overgrowth rims of zircon in migmatitic gneiss and biotite-sillimanite gneiss yielded weighted mean 207Pb/206Pb ages of 1861±4Ma and 1860±9Ma, respectively, coeval with the melt crystallization in leucosomes (e.g., [101103]). Such a timeline for partial melting and melt crystallization is consistent with that determined from the southern Yeongnam Massif, where anorthositic-gabbroic magmatism and partial melting are well constrained at ~1.87–1.86 Ga [33]. In particular, mafic rocks of the AMCG suite were emplaced diachronously at 1870±2Ma and 1861±6Ma, respectively [32]. In contrast to migmatitic gneisses, the fingerprint of ~1.86 Ga anatexis is lacking in the Buncheon granitic gneiss, except for an array of apparently concordant zircons ranging from ~1.95 to 1.90 Ga (Figure 10(d)); similar features were also documented by Kim et al. [42]. This scatter in zircon ages is attributed to the relatively dry nature of the Buncheon gneiss during the ~1.86 Ga melting event, yielding incomplete Pb loss or thermal annealing of zircons crystallized at 1981±2Ma. Thus, we suggest that the igneous protolith of amphibolite was emplaced at ~1866 Ma and subsequently metamorphosed at ~1862 Ma to yield partial melts in the presence of excess fluid. Regional ductile deformation likely ceased by 1852±3Ma, corresponding to the crystallization age of pegmatitic dyke. Taken together, the protolith of the amphibolite was emplaced, solidified, and then affected by synkinematic fluid-present melting during a relatively short time interval at the terminal stage of a Paleoproterozoic hot orogeny.

7.3. Petrogenesis of Tonalitic Leucosome

Tonalitic leucosomes produced by partial melting of mafic igneous rocks are common in high-temperature orogens, and such an anatexis may facilitate the segregation of newly formed felsic melts into the upper crust, significantly contributing to crustal differentiation [5, 27, 104]. Experimental studies (e.g., [11, 14, 15, 58, 105]) and phase equilibrium modeling (e.g., [13, 56, 106, 107]) suggested that melt production in mafic rocks is a continuous process and melt compositions vary from granitic to dioritic, depending on temperature, bulk composition, and the presence of excess fluid. Beard and Lofgren [15] experimentally investigated the dehydration melting of amphibolite at 800–1000°C and 1–6 kbar that yielded granodioritic to trondhjemitic melts and restitic assemblages of clinopyroxene+orthopyroxene+plagioclase+Fe–Ti oxides; in contrast, H2O-saturated melting yielded peraluminous, low-Fe melts together with amphibole-rich, plagioclase-poor residues. The latter is compatible with our petrographic observation (Figures 3 and 4) and Fe-poor, Ca-rich, and aluminous composition of tonalitic leucosomes (Figure 6). Based on thermodynamic modeling of open-system melting in amphibolites, Stuck and Diener [13] suggested that the compositions of melts produced by hornblende breakdown progressively change from tonalite–trondhjemite to granite fields with a decreasing temperature in the An-Ab-Or diagram (Figure 7). This result, however, is in marked contrast to the tonalitic composition of leucosome in the Okbang amphibolite (Figure 7).

The leucosome may not always represent primary melt compositions because they are readily modified by fractional crystallization and/or melt loss. In particular, plagioclase fractionation is one of key factors for controlling melt compositions [24]. The REE patterns of the analyzed leucosomes show positive Eu anomalies [Eu/EuN=1.546.88; Figure 8(b)], indicating that plagioclase with high Eu partitioning coefficient [108] has played a significant role during melt crystallization, locally yielding plagioclase-rich leucosomes (Figure 5(g)). Nevertheless, limited scatters in Th, U, Hf, and Zr concentrations may reflect the retention of U-bearing minerals such as zircon during melt crystallization [22]. In addition, the majority of leucosomes within the amphibolite are apparently confined to the amphibolite source area, and they show a limited range of tonalitic compositions (Figure 7). Thus, the compositions of tonalitic leucosomes in the Okbang amphibolite probably represent those of initial melts although melt extraction and migration are locally facilitated by regional deformation coeval with the melting event (e.g., [109]).

Based on the geochemical data of tonalite and amphibolite, Yakymchuk et al. [24] recently suggested that the fluid-present melting of amphibolite generally produced tonalitic leucosome as initial melt, and the paucity of K-feldspar and biotite in mafic source rocks is a critical factor for controlling melt composition. This type of melting is operative in the Okbang amphibolite where K-feldspar is generally lacking. In contrast, Stuck and Diener [13] suggested that initial solidus melting of amphibolite, accompanied by the biotite breakdown, generally produced K-rich granitic melts, but the hornblende dehydration at higher temperatures yielded granodioritic to tonalitic melts. These results are corroborated by a progressive change in melt compositions of metamafic rocks in the Ivrea Zone [110]. The variability in melt composition, however, is inconsistent with tonalitic melt compositions prevalent for the fluid-present melting of mafic rocks, including the Okbang amphibolite (Figure 7).

The pathway of external fluid is another key element for understanding the process of H2O-fluxed melting in the orogenic belt. Regional faults or shear zones commonly play a major role as fluid channels during fluid-present melting in migmatite complexes [25, 26, 111113]. In the Okbang amphibolite, the amount of leucosomes apparently increases towards ductile shear zones, suggesting that high-strain zones afforded pathways for transporting aqueous fluid and melt (Figures 3 and 4). Synkinematic fluid-induced melting is corroborated by the leucosome distribution typified by a network of one or more sets of extensional shear bands and boudin necks (Figures 3(e), 4(a), and 4(b); [44, 109]). On the other hand, open-system melting induced by an ingress of external fluid or melt could be recorded in the Hf isotopic compositions of zircon [114, 115]. 176Hf/177Hf ratios and εHft values of zircon in the leucosome are lower and more variable than those in the amphibolite (Figure 11; Table DR2), suggesting that the influx of external Hf component in migrated melts is significant in the leucosome [116, 117]. Thus, Hf isotopic values in the leucosome are interpreted to reflect the contribution of external melts derived from the surrounding metasedimentary rocks whose εHft values are in the range of –11.9 to –6.1 [35]. Positive zircon εHft values (4.2 to 6.0) in the amphibolite further suggest that its parental magma is most likely derived from juvenile mantle-derived source, confirming the previous result based on enriched midocean ridge basalt geochemistry and positive bulk-rock εNdt values [28, 29].

7.4. Implications for Tectonic Evolution of the Yeongnam Massif

Orogenesis in the continental crust generally comprises two stages: initial crustal thickening and subsequent gravitational collapse of the thickened crust [27, 118, 119]. During the latter extensional stage, upwelling asthenosphere may serve as a heat source for high-temperature metamorphism and anatexis (e.g., [120122]). Such a linkage could be represented by close association of mafic magmatism with the granulite-facies metamorphism in the Yeongnam Massif, as illustrated in a conceptual geodynamic model of Figure 12 [31]; arc-related magmatism and collisional orogeny at ~2.0–1.87 Ga are followed by late-orogenic mafic magmatism and regional metamorphism at ~1.87–1.85 Ga [30, 31, 33, 34]. The protracted orogenic history of the Yeongnam Massif is compatible with that of the North China Craton [37, 123]. In particular, the tectonothermal event at ~1.87–1.85 Ga is most likely linked to late-orogenic amalgamation between the Paleoproterozoic Korean arc and the North China Craton, ultimately leading to the formation of the Columbia/Nuna supercontinent [30]. Recently, Lee et al. [33] suggested that high-temperature low-pressure metamorphism is associated with midcrustal emplacement of the AMCG magma at ~1.87–1.86 Ga during an extensional regime in the southern Yeongnam Massif. Such a tectonothermal event is consistent with that found in the Okbang amphibolite and host metasedimentary rocks ([28, 34, 41]; this study). Our study further revealed that the Okbang amphibolite has experienced fluid-present melting shortly after its protolith emplacement, suggesting that the late-orogenic process in the Yeongnam Massif has culminated at ~1862 Ma. Thus, we conclude that the entire massif has concurrently experienced widespread magmatism and high-temperature metamorphism at ~1.87–1.86 Ga during the waning stage of Paleoproterozoic hot orogenesis (Figure 9(c); [33, 124]).

Fluid-present melting of mafic crustal fragment in the Yeongnam Massif ([33]; this study) sheds light on the late-orogenic evolution of Paleoproterozoic Korean arc situated along the trailing edge of the North China Craton [30, 124]. In particular, such a melting at upper amphibolite-facies condition may produce a large amount of melt to be retained in host rocks because of negative volume change of the melt-forming reaction [26]. The presence of partially molten rocks should result in the weakening of orogenic hinterland to trigger the onset of extensional tectonics [125]. Thus, fluid-present melting in the Okbang amphibolite is most likely linked to the late-orogenic extensional event at ~1.86 Ga prevalent in the Yeongnam Massif [31, 33].

The Hf isotopic data of zircon are useful for deciphering tectonic processes involved in the growth and recycling of continental crust [126]. At the onset of supercontinent assembly, subduction-related processes may lead to the generation of juvenile magmas with enhanced radiogenic Hf signature, affected by the mixing of variable proportion of recycled crustal materials [127]. During back-arc closure, crustal thickening and reworking dominate to yield negative εHft of magmatic zircons. On the other hand, orogenic collapse following the crustal accretion could be associated with mantle upwelling to yield juvenile Hf isotopic signatures [128]. Such a scenario for geodynamic evolution from subduction to accretion is compatible with the variation in Hf isotopic signatures of the Yeongnam Massif: εHft values at ~2.0–1.9 Ga range from –12.0 to –1.0 except for one zircon, whereas those at ~1.9–1.85 Ga from –8.0 to 6.0 (Figure 11; [31, 35]). Positive values in the latter are indicative of juvenile mafic magmas generated in an extensional setting, as revealed by the Okbang amphibolite as well as the Sancheong-Hadong AMCG suite [31]. In addition, zircon εHft values of basement gneisses in the Yeongnam Massif generally decrease with time at ~2.5–2.0 Ga. These variations during the Paleoproterozoic are consistent with the overall variation in zircon U-Pb ages and εHft values recorded in igneous rocks of the North China Craton, including the Damiao gabbro-anorthosite suite (Figure 11; [129134]). The decrease in εHft with time most likely reflects the increase in contribution of crustal recycling during the magma generation and ascent [135]. Moreover, ~1.9–1.8 Ga zircons in the North China Craton, including the Yeongnam Massif, are characterized by a large εHft spread (–17.0 to 9.0; Figure 11) but many zircons have positive εHft values, suggesting a significant influx of mantle-derived juvenile magmas. Such an εHft excursion is attributed to the emplacement of late- to postorogenic juvenile magmas, represented by the AMCG suites [31, 131, 136]. Taken together, the geodynamic and tectonic evolution of the Yeongnam Massif is closely linked to the Paleoproterozoic terrane assembly in the North China Craton, typified by prolonged accretionary–collisional orogenesis in association with final assembly of the Columbia/Nuna supercontinent [37, 123, 126].

The Okbang amphibolite in the Yeongnam Massif provides an excellent opportunity for understanding the fluid-present melting of mafic rocks and the role of fluid during the evolution of high-temperature terranes at the periphery of the North China Craton. The partial melting of amphibolitic protolith occurred at ~650–730°C and 4.6–5.2 kbar in the presence of externally derived fluid, particularly along a shear zone, and yielded tonalitic leucosomes containing peritectic hornblende megacrysts. The leucosomes are commonly distributed along axial plane of folds and within the interboudin partition of amphibolite, suggesting a synkinematic melting. Various microstructures including melt pockets and interstitial melts with rounded grains of hornblende or plagioclase in neosomes are indicative of in situ partial melting and melt crystallization. Our high-precision SHRIMP U-Pb zircon ages indicate that high-temperature metamorphism and anatexis of the amphibolite occurred at ~1861 Ma shortly after its formation at ~1866 Ma. Such a timeline for partial melting is consistent with 1862±2Ma crystallization of the leucosome and 1862–1861 Ma zircon overgrowth in migmatitic and biotite-sillimanite gneisses. The terminal phase of this melting episode is constrained by the intrusion of tungsten ore-bearing pegmatites at ~1852 Ma. All of these results are consistent with tectonothermal episodes at ~1.87–1.85 Ga recorded throughout the Yeongnam Massif (Figure 12; [30, 33]). Finally, the Okbang amphibolite has positive εHft of zircon, 4.2–6.0, suggesting that the parental magma was derived from a mantle source. Together with the Hf isotopic data available in the literature, the generation of such mafic magmas as well as high-temperature metamorphism in the Yeongnam Massif is compatible with the late-orogenic extensional phase of prolonged hot orogenesis, linked to the Paleoproterozoic crustal buildup of the eastern North China Craton.

All data generated or analyzed during this study are included in this published article (and its supplementary information files).

The authors declare that they have no conflicts of interest.

We thank C. Yakymchuk, an anonymous reviewer, and the editor S. Roeske for insightful and constructive comments that significantly improved the quality of this manuscript. We also acknowledge the help of K. Yi and Y.-J. Jeong for the SHRIMP and LA-MC-ICPMS analyses, respectively. This research was supported by the National Research Foundation 2016R1A6A3A01008112 and 2019R1F1A1055852 and the Basic Research Project (18-3111-1) of the Korea Institute of Geoscience and Mineral Resources (KIGAM) funded by the Ministry of Science and ICT (Information, communication, and technology) to Y. Lee.

Exclusive Licensee GeoScienceWorld. Distributed under a Creative Commons Attribution License (CC BY 4.0).

Supplementary data