Permian – Triassic adakitic igneous activity at Northern Mongolia: Implication for Permian – Triassic subduction system at the Siberian continental margin

The geochemistry of the Permian – Triassic large-scale igneous rock body of northern Mongolia is a key factor in understanding the subduction-related magmatism at the margin of “ Siberian continent. ” Several studies have been done in the Permian – Triassic igneous body; however, its detailed magmagenesis and tectonic significance remain unclear. This paper investigates the geochemistry of the Upper Permian andesites (Bugat/Baruunburen Formation) and the Late Permian – Middle Triassic plutonic rocks (Selenge plutonic rock complex) of the Per- mian – Triassic igneous body


Introduction
The Central Asian Orogenic belt (CAOB), which lies among Siberian craton, North China block, Tarim block, and East European craton (Sengӧr et al., 1993) is a crucial geologic unit in understanding the development process of the Eurasian continent (Kovalenko et al., 2004) (Fig. 1). It is generally accepted among scholars that the CAOB has been formed due to several processes such as subduction-accretion of the oceanic plate, volcanic arc magmatism, and collisions of continental fragments during the amalgamation of these cratons and blocks. In an area between the Siberian craton and the North China block, the geological setting of Mongolia is a significant factor to reveal the Paleozoic-Mesozoic tectonics of the CAOB (Fig. 1). The northern Mongolia, southern margin of the "Siberian continent" (Siberian craton + accreted geologic units due to Pre-Permian orogeny), widely exposes Precambrian-Paleozoic basement rocks of the Sayan-Baikal belt and Late Baruunburen Formation, and Khanui Group) and plutonic rock complexes (i.e., Permian-Triassic Selenge and Erdenet plutonic rock complexes) (Kepezhinskas and Luchitsky, 1974;Mossakovsky and Tomurtogoo, 1976;Sotnikov et al., 1995Sotnikov et al., , 2005Izawa and Ohsawa, 2004;Gerel and Munkhtsengel, 2005;Munkhtsengel, 2007;Munkhtsengel et al., 2007;Berzina et al., 2009;Tsukada et al., 2018a;Tsukada et al., 2021) (Fig. 3). The plutonic rock complexes, embedding substantial quantities of metal ores, have been well-studied mainly from the viewpoint of metallogeny (Khasin et al., 1977;Koval et al., 1989;Gavrilova and Maksimyuk, 1990;Jargalsaihan et al., 1996;Lamb and Cox, 1998;Watanabe and Stein, 2000;Dejidmaa et al., 2002;Berzina et al., 2005). However, no established theory seems to be there to explain the magmagenesis of the Permian-Triassic igneous rocks, which offers a key to understand the ancient arc system along the "Siberian continental margin." Morozumi (2003) suggested that the Selenge plutonic rock complex has an adakitic nature in the Sr/Y ratio. In addition, Munkhtsengel et al. (2007) mentioned that the Selenge plutonic rock complex corresponds to adakitic rocks formed at an arc condition. Tsukada et al. (2018a) found that the Upper Permian volcanic rocks of the Tulbur Formation of Khanui Group are adakitic rocks derived from the oceanic slab-melt. As mentioned above, several studies in Permian-Triassic magmatism have been made on the Khanui Group and the Selenge plutonic rock complex; on the other hand, little attention has been given to the Mogod and Bugat/Baruunburen Formations. In order to clarify the Permian-Triassic magmatism along the "Siberian continental margin," geochemical investigation of these crucial formations and comparison with the other volcano-plutonic rocks are necessary. This paper describes the lithology and geochemistry of the Bugat/Baruunburen Formation at the Jargalant area, Selenge plutonic rock complex at the Erdenet north area, and intermediate dike intruding into these rocks around the Erdenet city, northern Mongolia. It discusses the magmagenesis and subduction system along the Permian-Triassic Siberia-South Mongolian-Bureya continental margin.

Lithology and structure of the Bugat/Baruunburen Formation at Jargalant area
The study area, Jargalant village, 20 km east of Erdenet city, Mongolia, exposes intermediate volcanic rocks of the Bugat/Baruunburen Formation (Amgalan et al., 2017) (Figs. 4, 5). The rocks of this area, striking northeast and gently dipping south, are further divided into the following two parts in ascending order: (1) lower part (intermediate lava with minor tuff breccia) and (2) upper part (intermediate tuff and tuff breccia with lava intercalations) (Fig. 5).
The lava in the lower and upper parts is fine-to coarse-grained and vesicular, and has porphyritic or trachytoid textures and is holocrystalline in places (Fig. 6a, b). The vesicles, some of which are filled with calcite or chlorite to form amygdules, are elongated to oval shape. The platy joint strikes northeast and dips south by 20-50 • , and the   vesicles plunge southeast by 0-60 • . The lava is partly altered to the extent that some minerals have been replaced by calcite, chlorite, and opaque minerals such as hematite. In the lava showing porphyritic or trachytoid textures, subhedral or euhedral plagioclase, clinopyroxene (Cpx), and hornblende, more than 0.5 mm in major axis, lie in a groundmass composed of smaller plagioclase and interstitial chloritic material ( Fig. 7a, b). Later overgrowth and entrapped inclusions are not recognized at each mineral (Fig. 7b). The coarse lava rarely includes subhedral/euhedral zircon. The mineral assemblage and texture of the lava are monotonous throughout the study area. The tuff breccia in the upper part includes abundant angular clasts of intermediate volcanic rocks, up to 10 cm in diameter, in a cognate matrix (Fig. 6c). The tuff breccia has clear bedding, striking northeast-southwest, and dips south by 15 • (Fig. 6d).
The rocks of the Bugat/Baruunburen Formation are intruded by intermediate dikes (Fig. 6e). The dikes, having clear chilled margin in some places, generally strike northeast and are less than 100 m wide. Numerous veins of calcite, epidote, and chlorite are, in places, developed around the dike. The dikes are composed of subhedral or euhedral plagioclase, Cpx with porphyritic or trachytoid textures same as the lava (Fig. 7c, d). Later overgrowth and entrapped inclusions are not recognized at each mineral (Fig. 7d).

Lithology of the Selenge plutonic rock complex at Erdenet north area
The study area broadly exposes medium-grained granodiorite and diorite with a subordinate amount of andesite in the Selenge plutonic rock complex (Fig. 8). The andesite gradually changes into the diorite and granodiorite to show no distinct boundary. The rocks are mainly composed of euhedral or subhedral plagioclase, potassium feldspar, quartz, and biotite with subordinate amounts of apatite, zircon, and opaque minerals. The granodiorite and diorite commonly include mafic enclaves and sub-vertical joints developed in places ( Fig. 6f, g).
The rocks of the Selenge plutonic rock complex are intruded by intermediate dikes of andesite and diorite with a clear chilled margin (Fig. 6g, h). The intermediate dike includes the xenolith of granodiorite. The dikes are the mostly same, in mineral assemblage and texture, to that intrudes into the Bugat/Baruunburen Formation.

Whole rock chemical composition
Twenty-two samples of lava from the Bugat/Baruunburen Formation, 21 samples of plutonic rocks from the Selenge plutonic rock complex, and 26 samples from the intermediate dike were examined. Major element composition was determined using X-ray fluorescence (XRF; Rigaku Primus II ZSX equipped with Rh X-ray tube, 50 kV, 60 mA) installed at Nagoya University of Japan and Mongolian University of Science and Technology (MUST). Some trace elements (i.e., Co, Rb, Y, Zr, Nb, and Th) of the Selenge plutonic rock complex and intermediate dike were determined using XRF at MUST. In the XRF analysis, glass beads were prepared by fusing mixtures of 0.5 g of powdered sample with 5.0 g of lithium tetraborate for the major elements and mixtures of 1.5 g of powdered sample with 6.0 g of lithium tetraborate for the trace elements. Calibration was carried out using standard rock samples issued by the Geological Survey of Japan. Analytical precision of major elements was estimated to be < 1% for Si and about 3% for other elements, except for CaO, MgO, and Na 2 O, whose analytical precision is > 3% when the measured level is < 0.1% (Takebe and Yamamoto, 2003). Trace elements and rare earth elements (REE) were analyzed using inductively coupled plasma mass spectrometry (quadrupole type ICP-MS; Agilent 7700x) installed at Nagoya University, with the method described in Yamamoto et al. (2005). About 30 mg of each sample was digested with a mixed solution of HF-HClO 4 (2:1 by volume) at 150 ℃. After complete evaporation of the acids, 2 ml of 1.7 N-HCl was added to dissolve the cake. The residue was separated by centrifugation at 12, 000 rpm with a 2 ml polypropylene tube. The supernatant after centrifugation was transferred to another 10 ml Teflon beaker. The residue was then fused with HF-HClO 4 (2:1 by volume) again at 150 ℃. The fused cake was dissolved with about 2 ml 1.7 N-HCl by mild heating, and the solution was centrifuged at 12,000 rpm. In most cases, no residue was recognized after centrifugation. The HCl solution was evaporated to dryness. The fused cake was re-dissolved in 2% HNO 3 solution and determined by ICP-MS. In and Bi were used to trace ICP sensitivities; the In and Bi concentrations were mainly the same throughout the analysis. The oxide generation factor (LnO/Ln) was determined for each 20-ppb solution and used for REE analytical data correction. In the ICP-MS analysis, the correlation coefficients (R-value) of each element, calculated for five standard samples, were > 0.9994. The concentration relative standard deviation of the data was mostly less than 3%. The whole-rock composition of the samples is listed in Tables 2, 3, and 4. It is displayed on variation diagrams for selected elements against SiO 2 (Fig. 9).

Clinopyroxene chemical composition
The Cpx chemical composition in the samples was determined with a method described by Tsukada (2018) using energy-dispersive X-ray spectrometer (EDX, Oxford X-Max) linked with scanning electron microscope (SEM, Hitachi S-3400 N) at Nagoya University. The total values of each analysis were normalized as 100%, following Tsukada (2018). The analytical precision was estimated to be < 5% when the measured level was more than 1 wt%, and < 10% when the measured level was less than 1 wt% (Tsukada, 2018). Elements with a concentration of less than 0.5 wt% may not have been detected (Tsukada, 2018).
No.: number, FeO* : total iron as FeO, -: not analyzed. Trace elements analyzed using XRF are shown with bold italic.

Discrimination of the samples
The chemical composition of volcanic rocks gives evidence for the tectonic setting of the volcanic activity, and many diagrams to discriminate igneous rocks have been proposed (e.g., Miyashiro, 1974;Pearce, 1982). In this section, discrimination diagrams are used to discuss the tectonic setting of the rocks from the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike.
Data from the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike are arranged in a straight line in the variation diagrams for many elements. It likely suggests that these three rocks originated from a magma source through a fractionation process, or products as a result of magma-mixing between mafic and felsic magmas (Fig. 9).
The spidergrams of the samples show geochemical features similar to those of arc-related igneous rocks, such as enriched large ion lithophile elements and LREE in comparison to high-field strength elements and HREE with a negative Nb anomaly (Gill, 1981;Pearce et al., 2005) (Fig. 10). The majority of the samples are plotted in the calc-alkaline field in the SiO 2 vs. FeO* /MgO diagram (Miyashiro, 1974) (Fig. 16a). Taking all these lines of evidence into consideration, it can be concluded that the samples are of calc-alkaline rocks, which were formed in a volcanic arc environment. Extremely high Sr concentration and Sr/Y ratio of the samples correspond to "adakites" (Defant and Drummond, 1990). The adakites are generally characterized by SiO 2 > 56 wt%, Al 2 O 3 > 15 wt%, and Sr > 400 ppm, a high Sr/Y ratio, and fractionated REE (Defant and Drummond, 1990;Drummond et al., 1996;Martin, 1999). The samples in this study, SiO 2 50-69 wt% (avg. 56 wt%), Al 2 O 3 15-20 wt% (avg. 17 wt%), Sr 574-2572 ppm (avg. 1379 ppm), Sr/Y 44-212 (avg. 96), and La/Yb 13-100 (avg. 35), have adakitic nature in this aspect. It is supported by examination using Sr/Y vs. Y and La/Yb vs. Yb diagrams (Defant and Drummond, 1990;Defant et al., 1991;Moyen, 2009) (Fig. 16b, c). In the (K 2 O + Na 2 O) vs. SiO 2 diagram, most of data are plotted within the field of the adakite (Zhang et al. 2021) (Fig. 12). Martin (1999) classified adakites into the high silica adakite (HSA) with SiO 2 > 60 wt% and the low silica adakite (LSA) with SiO 2 < 60 wt%, and described that their geochemical features are different each other. Many samples of this study are assigned to Martin's LSA in terms of SiO 2 < 60 wt%, and most samples are plotted across the LSA and HSA fields in the discrimination diagrams (Martin, 1999) (Fig. 17). In the Cr/Ni vs. TiO 2 diagram, although the most samples are fall on the LSA field, the trend plots showing is similar to HSA's one (Fig. 17). Zhang et al. (2021) mentioned that the compositional range of the HSA given by Martin et al. (2005) falls within the initial definition of adakites established by Defant and Drummond (1990), whereas LSA can be termed adakitic rocks, but not strictly adakites. Following Zhang et al. (2021), the examined samples here are called to the "adakitic rocks."

Magmagenesis of the examined rocks
It is suggested that adakitic magma is generated by partial melting of the subducted oceanic slab (Defant and Drummond, 1990;Rapp, 1995;Sen and Dunn, 1994;Tsuchiya et al., 2007), and various processes during its ascent led to the diversity of the adakitic rocks (Castillo, 2006; Tsuchiya, 2008;Moyen, 2009;Zhang et al., 2021). Besides, an alternative way of the adakitic high Sr/Y and La/Y magma generation is proposed as magma-mixing, partial melting of the basaltic continental lower crust, and others (Atherton and Petford, 1993;Guo et al., 2007;Streck et al., 2007;Zhang et al., 2013;Zhang et al., 2021). It is known that an adakitic signature in rocks is produce by some magma-mixing processes at continental crust (Guo et al., 2007;Streck et al., 2007;Zhang et al., 2013), and the liner arrangement of the data from the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike in the variation diagrams might implies such magma-mixing. Guo et al. (2007) considered that the adakitic melts caused by various degrees of partial melting were generated in the lower crust, and the mixing of potassic-ultrapotassic magma with these melts produced a variety of adakitic rocks in southern Tibet after the collision of Indian and Eurasian continents. The trace elements compositions and Sr-Nd-Pb isotope characteristics of the Tibetan adakite are in close agreement with those of both the melt resulting from partial melting of the subducted oceanic plate and that from partial melting of the lower crustal mafic-intermediate rocks. And, Guo et al. (2007) considered the latter as a candidate for the source of the Tibetan adakite on the ground that the subducted oceanic plate was perhaps too cold to cause a melt at the end stage of continent-continent collision. They also mentioned that the heat required to melt the rocks of the lower continental crust could have been supplied by the contemporaneous potassic-ultrapotassic magmas combined with crustal heating induced by lithospheric extension. In the case of Mongolia, unlike the case of Tibet, the Mongol-Okhotsk oceanic plate is considered to have subducted beneath the "Siberian continental margin" in the Late Paleozoic (Kurihara et al., 2008;Bussien et al., 2011;Gordienko et al., 2012;Takeuchi et al., 2012;Onon and Tsukada, 2017). Although there is no concrete data on the stress field of the northern Mongolia around the Permian-Triassic, a compressional field, rather than extensional one, would be expected from the tectonic situation. Indeed, there is no evidence here for lithospheric extension, for example regional normal fault system, graben deposits, back-arc volcanism etc., that would have allowed thermal intrusion into the lower continental crust. Thus, it might not be possible to apply the model proposed by Guo et al. (2007) to the Mongolian case. Streck et al. (2007) concluded that the adakite-like high-Mg andesite at the Mount Shasta, western coast of US, was formed by mixing of dacitic and basaltic magmas and entrainment of ultramafic crystal material, along with the evidences such as orthopyroxene with high-Mg# rim and entrapped glass inclusions in olivine. And, Zhang et al. (2013) suggested that the adakite-like high-Mg diorites in eastern Dabie orogen, East China is a product resulted as a magma-mixing between the crust-derived granodioritic magma and the differentiation products of mantle-derived gabbronoritic magma with the following evidences. (1) the gabbronorite is usually occurred as enclaves in the high-Mg diorite host, and these are gradually changed each other without sharp boundary.
(2) plagioclase phenocrysts in the high-Mg diorite show a strong compositional zoning, i.e. Ca-rich core and Ca-poor rim. (3) The amphibole and biotite in the high-Mg diorite have distinctively higher F concentrations than those in the gabbronorite implying that the high-Mg diorite is a result of mixing of F-rich granodioritic magma and F-poor gabbronoritic magma. (4) In the high-Mg diorites, Cpx is included in amphibole to suggest a reaction between pre-existing Cpx and surrounding hydrous melts causing the amphibole formation. (5) Zircons from the gabbronorite and the diorite are different each other in their morphology, internal texture, Hf isotope, and trace element composition.
In the case of this study, evidences positively suggesting the magmamixing, such as compositional zoning and entrapped glass inclusions in Fig. 10. MORB-normalized multi-element concentration diagrams (spidergram) of the examined samples. Normalizing MORB composition by Sun and Mcdonough (1989) is used. The data of the Tulbur Formation are from Tsukada et al. (2018a). The range of the Kitakami adakitic granites is from Tsuchiya et al. (2005). crystals, are not recognized at least in microscopic and SEM observations for 46 samples (Figs. 7,13,and 14,Table 5). Besides, monotonous chemical composition within Cpx crystals also support this view (Figs. 14 and 15). The intermediate dikes always intrude into the host lava and plutonic rocks with clear boundary, and not appear to have been reacted each other (Fig. 6e, g, and h). Therefore, it is considered Fig. 11. Chondrite-normalized REE patterns of the examined samples. Normalizing chondrite values are after Sun and McDonough (1989) and Yamamoto et al. (2005). The data of the Tulbur Formation are from Tsukada et al. (2018a).  (Irvine and Baragar, 1971). The data of the Tulbur Formation are from Tsukada et al. (2018a). The range of adakite was from Zhang et al. (2021).
unlikely, here, that the samples in this study are the result of magmamixing. Zhang et al. (2005) classified the adakitic rocks according to their origins into Type 1 (derived from the partial melting of the subducted oceanic slab) and Type 2 (derived from the partial melting of the basaltic lower continental crust). Type 2 adakitic rocks have lower Al 2 O 3 concentration (less than about 16 wt%), higher K 2 O/Na 2 O ratio (about 0.5 or more) than Type 1 reflecting their source composition: the basaltic rocks of the lower continental crust generally have a higher K 2 O/Na 2 O ratio than the oceanic basalt. An examination using K 2 O/Na 2 O vs. Al 2 O 3 diagram highly suggests that the present samples, low K 2 O/Na 2 O ratio (mostly less than 0.6) and high Al 2 O 3 concentration (15-20 wt%, avg. 17 wt%), correspond to the Type 1 adakitic rocks (Kamei et al., 2009;Liu et al., 2010) (Fig. 16d).
When an adakite magma are produced by various levels of partial melting of an eclogite or garnet amphibolite, a positive correlation would be expected for Sr/Y and La/Yb because Sr and La are incompatible whereas Y and Yb are compatible in garnet-bearing and plagioclase-free residues (e.g., Defant and Drummond, 1990;Rapp and Watson, 1995;Moyen, 2009;Liu et al., 2010) (Fig. 16e). Liu et al. (2010), based on geochemical studies of Cretaceous adakitic rocks in central-eastern China, demonstrated that adakitic rocks of oceanic slab-origin and those of lower continental crust-origin show quite different arrays on the Sr/Y vs. (La/Yb) N diagram, i.e. the former exhibits much higher Sr/Y value than the latter for a given (La/Yb) N value. Such a sharp trend of increasing Sr/Y ratio of oceanic slab-derived adakitic rocks is unlikely to be caused by plagioclase and garnet fractionation, and it probably reflects the low-temperature alteration of the oceanic slab (Liu et al., 2010). The most data here are plotted in the field of the slab-derived adakitic rocks in the Sr/Y vs. (La/Yb) N diagram, and therefore the examined rocks are quite likely of slab-melt origin (Fig. 16e).
The continental crust generally has a lower Ce/Pb than oceanic crust (Sun and McDonough, 1989;Taylor and McLennan, 1985;Rudnick and Gao, 2003). And the altered oceanic crust has a much higher Sr/La, than the lower continental crust, due to the LREE-depleted N-MORB composition and enrichment of Sr by seawater alteration (Liu et al., 2010). Therefore, the Sr/La of oceanic slab-derived adakitic rocks tends to be higher than that of lower continental crust-derived adakitic rocks. The present data, characterized by a comparatively wide range of Ce/Pb ratio and high Sr/La ratio, imply that partial melting of an altered oceanic slab and contamination of sediments on oceanic crust or continental crustal materials have played a role in the magma-formation (Liu et al., 2010) (Fig. 16f).
Some magma of adakitic rocks is considered to have been formed accompanying fractional crystallization (Macpherson et al., 2006;Jia et al., 2017). Macpherson et al. (2006) explained that the Pleistocene typical adakitic volcanic rock at the Mindanao Island, Philippines, was not attributed to the subducted slab-melting, but either was produced by fractional crystallization of garnet-bearing assemblage from basaltic arc magma, or was produced by melting of garnet-bearing solidified basalt. They pointed that an important feature of the Mindanao rock is the positive correlation between Dy/Yb and SiO 2 concentration, and chondrite-normalized REE pattern shows an increasing trend from medium to heavy REE in the rocks of which SiO 2 concentration is less than 60 wt%. The examined samples, unlike the Mindanao rocks, show negative correlation in Dy/Yb vs. SiO 2 diagram, and show a monotonous decreasing trend from light to heavy REE even in rocks with SiO 2 < 60 wt% at chondrite-normalized REE pattern (Figs. 11 and 18a). As the data of Mindanao and this study are quite different in behavior of REE, at present, it is more reasonable to suppose that the Mongolian adakitic initial melt was produced by slab-melting as described before, rather than by the fractional crystallization of basaltic arc magma/melting of solidified basalt as seen in Mindanao.
Incidentally, it is generally known that adakitic rocks are closely associated with porphyry copper deposits (Thieblemont et al., (1997); Sajona and Maury, 1998;Oyarzun et al. 2001;Ballard et al. 2002;Ishihara and Chappell, 2010;Liu et al., 2010;Sun et al. 2011Sun et al. , 2013. Thieblemont et al., (1997) noted that 38 of 43 large Au, Ag, Cu, and Mo deposits in the world are associated with adakitic rocks, and Leng et al. (2007) showed that most porphyry copper deposits in China are adakite-related. Ishihara and Chappell (2010) summarized the geochemical characteristics of copper-mineralized granitoids from Chile, Highland Valley (Canada), Erdenet (Mongolia), Dexing (China), Medet (Bulgaria), and Ani (Japan), and suggested that many of porphyry copper deposits are closely related to adakitic rocks originating from melt of oceanic slab which has undergone a hydrothermal alteration at oceanic ridge. They remarked that alteration by S-rich hydrothermal fluids in the ridge increases the S concentration in the comparatively copper-rich MORB, and the partial melting of such altered basalt favors the formation of large-scale porphyry copper deposits including copper sulfide minerals such as chalcocite, chalcopyrite, covellite. Furthermore, Liu et al. (2010) demonstrated, in comparison the Cretaceous ore-bearing and ore-barren high-Mg adakitic rocks in central-eastern China, that the deposits are characteristically occurred in adakitic rocks of slab-melt origin. In terms of whether mineralization is associated, the Permian-Triassic igneous rocks of Mongolia host one of the major porphyry copper deposits in the world, and this fact likely supports the oceanic slab-melt origin for the present samples. It is known that the Mg# in the melt generated by partial melting of MORB is up to 45, and interaction with peridotite increases the Mg# in the melt (Rapp, 1997;Rapp et al., 1999). The Mg# of the present samples is generally higher than that of the experimental slab-melt, and many of them are more than 45 (Rapp et al., 1991;Rapp and Watson, 1995;Sen and Dunn, 1994;Winther and Newton, 1991) (Fig. 18b). This fact points that the initial magma of the examined rocks interacted with mantle peridotite. That is, the magma of present samples has not been derived from basaltic lower crust-melt, but is highly possible to have been derived from the oceanic slab-melt interacted with mantle peridotite during its rising. The rocks in this study have a higher concentration in Fe, Mg, Cr, Co, and Ni than the adakitic granites of the Kitakami Mountains in Japan, which represent the primitive slab-melt composition (Tsuchiya et al., 1999) (Fig. 10). Such rocks have been reported from some regions, e.g. Aleutian-arc and Kitakami Mountains.
For example, Drummond et al. (1996) described Ni-enriched adakitic rocks from the Aleutian-arc, and they attributed it to the interaction between slab-melt and mantle peridotite. Furthermore, adakitic rocks with high Mg, Cr, and Ni concentration, similar to the rocks of this study, exposed at the Kitakami Mountains result from the interaction between the slab-melt and mantle peridotite (Tsuchiya et al., 1999(Tsuchiya et al., , 2005. The examined samples may correspond to the LSA of Martin (1999). Martin (1999) considered that the LSA was resulted from the melting of mantle peridotite that had been metasomatized by reaction with basaltic slab-melt. While, Tsuchiya et al. (2005) showed that the Ni/Cr ratio of the modified mantle-melt is smaller than that of the melt produced by slab-melt and mantle interaction. In the Ni-Cr correlation, the present data are not plotted in the field of modified mantle-melt, but appear at or near the mixing line of the mantle and slab melt (Fig. 18c). This suggests that the initial magma of the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and the intermediate dike were formed by slab-melting and then interacted with mantle peridotite during its ascent.
Nb and Ta are known to have same incompatibility each other, and are not fractioned during mantle magmatism, but fractioned during crustal processes (Sun and McDonough, 1989;Foley et al., 2002;Xiao et al., 2006). Therefore, Nb/Ta is likely a good indicator to know the magmatic origin and its contamination at the continental crust. Nb/Ta of the MORB and primitive mantle is almost constant in 17 (Sun and   McDonough, 1989). The Nb/Ta ratios of the upper and middle continental crusts are 13.4 and 16.5 respectively, and that of the lower crust is 8.3 (Nb/Ta: 12.4 at total crust in average) (Rudnick and Gao, 2003). Nb/Ta of the adakite and trondhjemite-tonalite-granodiorite (TTG) melt ranges from 1.5 to 20, and basically have lower Nb/Ta ratio than the N-MORB and primitive mantle (Sun and McDonough, 1989;Foley et al., 2002). The Zr/Hf ratios in the N-MORB, primitive mantle, and continental crust are nearly identical, ranging from 33 to 37 (Sun and McDonough, 1989;Foley et al., 2002). In the Nb/Ta vs. Zr/Hf plot, the present data are plotted along an array including N-MORB, primitive mantle, and continental crust within the Nb/Ta range of adakite and TTG (Fig. 18d). This may indicate that upwelling adakitic magma underwent various degrees of continental contamination.

Tectonic implications of the adakitic magmatism
The Bugat/Baruunburen Formation and Selenge plutonic rock complex rocks have a signature of adakitic rocks. Source magma was formed by an interaction between slab-melt and mantle peridotite. The Bugat/ Baruunburen Formation gives zircon U-Pb age of 257.9 ± 3.8 Ma and the Selenge plutonic rock complex gives 40 Ar-39 Ar ages of 247-258 Ma and SHRIMP ages of 243-250 Ma (Gerel and Munkhtsengel, 2005;Sotnikov et al., 2005;Munkhtsengel, 2007;Munkhtsengel et al., 2007;Tsukada et al., 2021) (Table 1). Besides, Tsukada et al. (2018a) reported that the intermediate lava of the Tulbur Formation of the Khanui Group, dated as 256.9 ± 2.2 Ma by zircon U-Pb method, has almost the same geochemical characteristics as the examined samples (Figs. 10,11,12,16,18). Furthermore, the rocks of the Khanui Group and Selenge plutonic rock complex exposed around Darkhan city, 250 km northeast of Bulgan city, show a signature of slab-melt-origin adakitic rocks (Tsukada et al., 2018b). Therefore, it is evident that adakitic magmatism caused by subducted slab-melting occurred at the "Siberian continental margin" around 250 Ma.
The partial melting of MORB requires a particular condition at 2.0 GPa and 900-950 ℃, and "hot and young (<5 Ma) slab subduction model" is considered to be the best explanation for the generation of Type 1 adakitic magma (Defant and Drummond, 1990;Sen and Dunn, 1994;Rapp and Watson, 1995;Tsuchiya et al., 2007). Onon and Tsukada (2017) deduced that the Khangai-Daur belt, a Late Paleozoic-Early Mesozoic accretionary complex, is the fossilized remains of the subducted Mongol-Okhotsk oceanic plate and its cover beneath the "Siberian continent." The migration time of oceanic plates from ridge/oceanic islands to trench can be estimated from the fossil record of radiolarian chert in the accretionary complex. The movement of the Mongol-Okhotsk oceanic plate from the oceanic island to the trench is inferred to have had taken more than 50 million years based on the radiolarian biostratigraphy of the chert in the Late Paleozoic accretionary complex of the Khangai-Daur belt (Kurihara et al., 2008;Tsukada et al., 2013;Bayart et al., 2022). The subducted oceanic plate underneath the "Siberian continent" at the Late Paleozoic time was old and cold when it reached the trench. Therefore, the generation of initial magma of the Permian-Triassic adakitic rocks was probably the result of melting of cold and old subducted slab, and the "partial melting of hot and young (<5 Ma) subducted slab" model is incompatible with this case.
Several specific settings, such as flat subduction and oblique subduction, have been proposed to account for the melting of the cold and old slab (Gutcher et al., 2000;Yogodzinski et al., 2001). Gutcher et al. (2000) demonstrated that the flat subduction which causes isobaric heating produces partial melting of the 50 million years old subducted slab, causing a broad adakitic rock belt of 100 km width or more. Permian-Triassic adakitic rocks have been reported from a wide area, covering 200 km × 50 km, including Burgan, Erdenet, and Darkhan in northern Mongolia (Morozumi, 2003;Munkhtsengel et al., 2007;Tsukada et al., 2018a, b) (Figs. 2 and 4). Although it might be possible that the Permian-Triassic igneous rocks correspond to a "broad adakitic rock belt" attributed to the flat subduction of the Mongol-Okhotsk oceanic plate, precise age and geochemical feature of the Permian-Triassic igneous rocks are, in many areas, still unclear. Further study on age and geochemistry at the western part of the Permian-Triassic igneous rocks is needed for clarification (Fig. 2). Yogodzinski et al. (2001) proposed that oblique subduction and slab tearing along an L-shaped trench result in the melting of cold and old slab according to studies in the Aleutian-Kamchatka region. According to them, oblique subduction and tearing of the slab lead to heat intrusion and slab-melting at the torn edge of the slab. The paleolatitude of the Upper Permian Bugat/Baruunburen Formation is calculated to be 37.1 • N based on thermal remanent magnetization data presented in Pruner (1987). According to the reconstructed Late Permian continental distribution, the adakitic volcanism had occurred along the L-shaped trench between the eastern margin of the Siberian continent and the northern margin of the South Mongolian-Bureya continent (Ueno, 2006;Metcalfe, 2011;Cleal, 2018) (Fig. 19a). The shape of the subduction zone along the Siberia-South Mongolian-Bureya continental margin is similar to that of the Aleutian-Kamchatka region. The magma generation of the Permian-Triassic adakitic rocks may also have been related to oblique subduction, like the Aleutian-Kamchatka case (Fig. 19). However, this remains to be confirmed when further data on relative plate motion between the continents and the Mongol-Okhotsk oceanic plate, e.g., moving path of the Mongol-Okhotsk oceanic plate based on paleomagnetic data, become available.
It is known that the superplume activity that formed the Siberian traps had taken place just behind the volcanic arc at Late Permian (Sobolev et al., 2011) (Fig. 19). Another possible scenario is that the heat supply from the superplume that formed the Siberian traps caused partial melting of the oceanic slab subducted beneath the continental margin (Fig. 19). At present, there is no clear evidence that the heat supply from the superplume played a role in the slab-melting, but it is unlikely that the mantle wedge was completely free from the thermal effects of superplume activity. It might be necessary to examine the cause of the adakitic igneous activity in northern Mongolia concerning the superplume in future studies. The intermediate dike has remarkably similar geochemical characteristics to the Bugat/Baruunburen Formation and Selenge plutonic complex (Figs. 10,11,12,16,18). This indicates that the adakitic magmatism had lasted after the Early Triassic. The Erdenet plutonic rock complex which is Middle-Late Triassic, the younger igneous rock unit than the Selenge plutonic rock complex, also shows adakitic nature . In addition, Donskaya et al. (2012) presented the Late Triassic adakitic volcanic rocks from the Kataev Formation at the western Transbaikalia, northeastern extension of the Permian-Triassic igneous rocks. The adakitic igneous activity caused the dike, therefore, likely corresponds to the post-Middle Triassic adakitic igneous activity such as the Erdenet plutonic rock complex and the Kataev Formation. The Late Triassic adakitic volcanic-plutonic rocks were also reported from the Erguna massif at northeast China and Middle Gobi volcanic belt of south Mongolia on the South Mongolian block (Li et al., 2017;Sheldrick et al., 2020;Tang et al., 2021) (Figs. 1,  19). Li et al. (2017) mentioned that the Middle Triassic volcanic rocks at the Erguna massif have an affinity to normal arc-type volcanic rocks, but  (Miyashiro, 1974). FeO* : total iron as FeO. The data of the Tulbur Formation are from Tsukada et al. (2018a). (b) Sr/Y vs. Y diagram (Defant and Drummond, 1990). ADR: andesite-dacite-rhyolite. Adakite and Island arc ADR fields are after Defant et al. (1991); Defant and Drummond (1990), and Martin et al. (2005). The data of the Tulbur Formation are from Tsukada et al. (2018a). (c) La/Yb vs. Yb diagram (Moyen, 2009). Data of the Tulbur Formation (Tsukada et al., 2018a) are also plotted. (d) K 2 O/Na 2 O vs. Al 2 O 3 diagram (Kamei et al., 2009(Kamei et al., , 2013Liu et al., 2010). (d) Sr/Y vs. La/Yb diagram (Liu et al., 2010). The data of the Tulbur Formation are after Tsukada et al. (2018a). OSA: oceanic slab-derived adakitic rocks in the central-eastern China. LCCA: lower continental crust-derived adakitic rocks at Dabie region, China. The OSA and LCCA fields are referred from Liu et al. (2010). (e) Ce/Pb vs. Sr/La diagram (Liu et al., 2010). The data of the Tulbur Formation are after Tsukada et al. (2018a). OSA: oceanic slab-derived adakitic rocks in the central-eastern China, LCCA: lower continental crust-derived adakitic rocks in the central-eastern China. The OSA and LCCA fields are after Liu et al. (2010).
the Late Triassic one is slab-melt-origin adakitic rocks. This might imply that the place of the melting of subducted oceanic plate spread from the Siberian continental margin to the South Mongolian-Bureya continental margin during the Middle to Late Triassic (Figs. 1, 19).

Conclusions
Geochemical investigations of the Bugat/Baruunburen Formation, the Selenge plutonic rock complex, and the intermediate dike were carried out in order to clarify the subduction system along the "Siberian continental margin" at Permian-Triassic. As a result, the following points were established.
i. Data from the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike forming a linear trend in the variation diagrams for many elements imply a single magmatic source for these three rocks. ii. The spidergrams of the examined samples show geochemical characteristics similar to the arc-related igneous rocks, and the majority are plotted in the calc-alkaline field in the SiO vs. FeO* / MgO diagrams. The conclusion is that the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike are calc-alkaline igneous rocks formed at a volcanic arc environment. iii. The samples have an extremely high Sr concentration, Sr/Y ratio, and fractionated REE, which gives evidence of adakitic rocks. iv. Low K 2 O/Na 2 O, high Sr/Y, high La/Yb, and high Sr/La ratios point that the magma of the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and the intermediate dike was derived from oceanic slab-melt. v. Higher concentration of Cr and Ni in the samples than the primitive slab-melt suggests that the magma of the Bugat/Baruunburen Formation, Selenge plutonic rock complex, and intermediate dike resulted from an interaction slab-melt and mantle peridotite. The high Mg# value in many samples supports this view. vi. The fact that Nb/Ta ratio of the present samples are plotted along an array including N-MORB mantle and continental crust likely suggests that the upwelling adakitic magma had undergone various degrees of contamination within the continental crust. vii. The paleolatitude of the Upper Permian Bugat/Baruunburen Formation is calculated to 37.1 • N. Therefore, the Late Permian-Middle Triassic adakitic igneous activity had taken place in the volcanic arc along the Siberian continental margin in the midlatitude region of the Northern Hemisphere. Fig. 17. Discrimination diagrams of high-silica and low-silica adakites (Martin, 1999). HAS: high-silica adakite, LSA: low-silica adakite.
viii. The intermediate dike having almost the same geochemical feature as the Bugat/Baruunburen Formation and Selenge plutonic complex is evidence that the adakitic magmatism lasted till after the Early Triassic.
It is generally considered that slab-melting occurs when a young/hot slab subducts beneath a continent. However, the oceanic plate subducted beneath the "Siberian continent" in the late Paleozoic is considered to be old and cold. Possible explanations for the old/cold slab-melting include flat subduction, slab tearing due to oblique subduction, and heat supply from the superplume that caused the Siberian traps. In order to clarify this issue, it is necessary to do further research on the movement direction of the Mongol-Okhotsk oceanic plate and the Permian-Triassic thermal structure beneath the "Siberian continent."

Declaration of Competing Interest
The authors declare that they have no known competing financial  Macpherson et al. (2006). (b) Mg#-SiO 2 relationship for the examined samples. The Field of the experimental slab-melt is after Rapp et al. (1991), Rapp and Watson (1995), Sen and Dunn (1994), and Winther and Newton (1991). The data of the Tulbur Formation are from Tsukada et al. (2018a). (c) Ni-Cr relationship for the examined samples. Data of the Tulbur Formation (Tsukada et al., 2018a) are also plotted. Ni and Cr compositions of the primitive mantle and inferred slab-melt are from Sun (1982) and Tsuchiya et al. (2005) respectively. Range of melt derived from modified mantle and inferred slab-melt are after Tsuchiya et al. (2005). (d) Nb/Ta vs. Zr/Hf diagram. LCC: lower continental crust, MCC: middle continental crust, UCC: upper continental crust, TTG: trondhjemite-tonalite-granodiorite. The data on N-MORB, primitive mantle, LCC, MCC, and UCC were from Sun and McDonough (1989) and Rudnick and Gao (2003). Nb/Ta upper and lower limits of adakite and TTG were from Foley et al. (2002).
interests or personal relationships that could have appeared to influence the work reported in this paper.  Ueno, 2006;Xiao et al., 2009;Metcalfe, 2011;Bazhenov et al., 2016;and Cleal, 2018) facilitating the use of the facilities in the university. We would like to thank Enago (www.enago.jp) for the English language review.
Part of the data was obtained through the Regional survey on the infrastructure development project for promoting rare metal resource exploration in 2013 (Mongolia) by the Ministry of Economy, Trade and Industry of Japan (METI). A part of this study was supported by the Daiko Foundation (Grant number: 9219), Japan.The second author, Mr. Nemekhbayar, P. passed away 1 July 2021 at the age of 32 years. This study was inspired by his master's study at Nagoya University on the geochemistry of the Tulbur Formation, Khanui Group, Mongolia.