Carboniferous magmatic records of central Mongolia and its implications for the southward subduction of the Mongol–Okhotsk Ocean

ABSTRACT The now-extinct Palaeozoic to Mesozoic Mongol–Okhotsk Ocean is evidenced by the Mongol–Okhotsk suture, which stretches from central Mongolia to the Sea of Okhotsk. The geodynamics of southward subduction of the Mongol–Okhotsk Ocean remain enigmatic, especially whether the subduction began during the Carboniferous is highly debated. In this paper we present new zircon U–Pb ages and whole-rock geochemical data for the Carboniferous magmatic rocks in Northeast Mongolia. Zircon U–Pb dating indicates that the Adaatsag dacite, Northern Bayanjargalan gabbro-diorite and Bagannur pluton were emplaced at ca. 325 Ma, ca. 316 Ma, and ca. 305 Ma, respectively. Occurrences of these rocks together constitute a Carboniferous magmatic belt to the northwest of the Middle Gobi volcanic-plutonic belt. The ages of trapped zircon xenocrysts within the Bagannur pluton indicate that the basement of this magmatic belt is the Ereendavaa terrane. Geochemical analysis indicates that these magmatic rocks were formed in an Andean-type active margin, which provides robust evidence for southward subduction of the Mongol-Okhotsk Ocean during the Late Carboniferous. By combining all the available data, we argue that the onset of the Central Mongolia–Tsagaan Uul continental collision during the Late Carboniferous – Early Permian may have triggered the beginning of an active margin of the Mongol-Okhotsk Ocean along the northern coast of the Central Mongolia microcontinent.


Introduction
The Central Asian Orogenic Belt (CAOB; e.g. Windley et al. 2007) or the Altaids (Şengör et al. 1993) is one of the largest and longest-lived accretionary orogenic collages in the world, which extends from the Uralides in the west to the Pacific margin in the east and is bounded by the Siberian Craton to the north and the Tarim-North China cratonic blocks to the south (e.g. Xiao et al. 2015) (Figure 1 (a)). The eastern CAOB was formed through the amalgamation of a variety of micro-continents, magmatic arcs, ophiolites, blueschists, fore-arc or back-arc formations driven by the evolution of the Paleo-Asian Ocean (e.g. Xu et al. 2015;Liu et al. 2017;Wang et al. 2017). The final closure of the eastern Paleo-Asian Ocean is proposed to occur either in the Devonian Zhao et al. 2016) or Permian-Triassic (Xiao et al. 2015;Huang et al. 2018;Liu et al. 2019). As the youngest segment of the eastern CAOB, the Mongol -Okhotsk orogenic belt extends more than 3,000 km in a northeastern direction from Central Mongolia to the Uda Gulf of the Okhotsk Sea (e.g. Zorin 1999;Bussien et al. 2011) (Figure 1(b)). This belt is proposed to be formed by the closure of the Mongol-Okhotsk Ocean (MOO), which is attested by the occurrences of ophiolites (Tomurtogoo et al. 2005;Zhu et al. 2018), sediments containing marine fossils (Halim et al. 1998), and intra-oceanic arcs (Zorin 1999). The final closure of the MOO is estimated to have occurred between the Early-Middle Jurassic and Early Cretaceous (e.g. Kravchinsky et al. 2002;Van der Voo et al. 2015;Demonterova et al. 2017;Sorokin et al. 2020;Yi and Meert 2020), which led to the collision of the Siberian Craton and the Amur Block (Tomurtogoo et al. 2005;Kelty et al. 2008;Tang et al. 2015). Thus, reconstruction of its evolution plays an important role in understanding of the amalgamation and growth of the eastern part of the Eurasian continent Ren et al. 2018).
As for the subduction of the MOO, the oceanic plate is proposed to have been subducted northward beneath the Siberian Craton and southward beneath the Amur Block (e.g. Metelkin et al. 2007;Donskaya et al. 2013;Tang et al. 2014;Zhao et al. 2017).
Northward subduction-related magmatism is represented by the Middle Carboniferous to Triassic Selenge arc and the Angar-Vitim granitoids in Transbaikalia (e.g. Mazukabzov et al. 2010;Larin et al. 2011;Donskaya et al. 2013). Southward subduction is evidenced by the Permian-Jurassic magmatic records in Central Mongolia, Erguna and Xing'an massifs (Li et al. 2013;Tang et al. 2016;Zhao et al. 2017;Liu et al. 2018). Moreover, southward subduction of the MOO could have facilitated convective erosion of the lithospheric mantle from upwelling asthenospheric flow in the North China Craton and southern Mongolia, which further constrain the timing and petrogenesis of the widespread East Asian Mesozoic-Cenozoic magmatism (Barry et al. 2017;Sheldrick et al., 2020 a, b). Detrital zircon data of the sandstones from the Middle Gobi volcanic belt, south of the Mongol-Okhotsk suture, imply that the southward subduction had begun during the Carboniferous (Bussien et al. 2011). However, the subduction-related Carboniferous magmatic records are lacking and understudied, which has hindered our understanding of the early subduction evolution of the MOO.
In this paper, we present new zircon U -Pb ages and whole -rock geochemical data for the Carboniferous magmatic rocks in Northeast Mongolia, to unravel their petrogenesis and to reconstruct geodynamic settings of their formation. These new results will shed light on the southward subduction evolution of the MOO during the Carboniferous.

Geological setting and sample description
The territory of Mongolia is subdivided into two tectonic domains by the Main Mongolian Lineament (MML), namely, the 'Caledonian' domain to the north and the 'Hercynian' domain to the south (Tomurtogoo 1997;Badarch et al. 2002). The northern domain is composed of Archaean-Proterozoic cratonic blocks, Neoproterozoic to Lower Palaeozoic metamorphic rocks and ophiolites, and Palaeozoic volcanic and sedimentary rocks. The southern domain mainly consists of Lower to Middle Palaeozoic ophiolites and arc-related volcanic rocks (Badarch et al. 2002;Windley et al. 2007). The northern domain is crosscut by the younger Mongol-Okhotsk Orogenic Belt (Bussien et al. 2011;Donskaya et al. 2013), resulting from the evolution of the MOO. A scissor-like oceanic closure model has been proposed for the MOO, which is reflected by the younging trend of sediments and intrusions from west to east along the orogenic belt (e.g. Zonenshain et al. 1990;Tomurtogoo et al. 2005;Metelkin et al. 2010;Donskaya et al. 2013;Tang et al. 2016). In the segment of Mongolia, the suture zone of the MOO is represented by the Adaatsag -Dochgol terrane, which are composed of schist, quartzite, metasandstone, phyllite, chert, metavolcanic rocks, coral limestone, and mélange containing fragments of ophiolite (Badarch et al. 2002;Tomurtogoo et al. 2005;Zhu et al. 2018). Zircon age data suggest that the Adaatsag ophiolite (ca. 325 Ma; Tomurtogoo et al. 2005) and Khuhu Davaa ophiolite (ca. 320 Ma; Zhu et al. 2018) in this zone were formed during the Carboniferous.
The Middle Gobi volcano-plutonic belt to the south of the suture zone could be the optimal study area for elucidating the history of the southward subduction of the MOO (Badarch et al. 2002;Tomurtogoo et al. 2005;Zhao et al. 2017). This belt extends northeast to the Erguna massif, where abundant Permian-Triassic intrusions and coeval porphyry-type ore deposits were formed due to the southward subduction of the MOO (Sun et al. 2013;Xu et al. 2013;Tang et al. 2014Tang et al. , 2016. According to Badarch et al. (2002), the basement of the Middle Gobi volcano-plutonic belt should be the Ereendavaa terrane, which mainly consists of Paleoproterozoic gneiss, amphibolite, schist and marble, Neoproterozoic black schist, metasandstone, limestone and minor conglomerate and volcanic rocks, unconformably overlain by Silurian -Devonian volcanic-sedimentary rocks (e.g. Narantsetseg et al. 2019). Though no Precambrian age is available for the Ereendavaa terrane (Miao et al. 2017), the granitic rocks/gneisses with Proterozoic ages have been reported in Erguna massif that is regarded as the extension of the Ereendavaa terrane (Tang et al. 2013;Shao et al. 2015).
Three magmatic complexes are sampled from central Mongolia to the south of the suture zone of the MOO ( Figure 2). Samples from dacite in the Adaatsag region are grey to dark purple and assigned to be Carboniferous-Permian age during the 1:500,000 geological mapping in the 1970s (Marinov et al. 1973). They exhibit a medium-grained porphyritic texture, and the modal abundances are plagioclase ~30 vol %, quartz ~10 vol %, magnetite ~5 vol %, and epidote 1-2 vol % with grain sizes between 0.1 and 0.6 mm, residing in a microlithic groundmass that is made up of plagioclase, quartz, and magnetite. Rutile and zircon are present as accessory minerals (Figure 3 (a, c)).
The gabbro-diorite from the Northern Bayanjargalan area intrudes a granitic pluton, and both of them were ascribed as the Cambrian during the 1:500,000 geological mapping. It is dark greygreen with a typical gabbroic texture (locally cumulate texture) and massive structure (Figure 3(b)). It consists primarily of hornblende (~40 vol%) clinopyroxene (~20 vol%), plagioclase (~35 vol%) and magnetite (~5 vol%). The hornblendes and clinopyroxene are euhedral, while the plagioclases are subhedral and weakly altered on the surface (Figure 3(d)).

Geochemistry
Whole-rock major and trace element compositions were determined at the Wuhan Sample Solution Analytical Technology Co., Ltd., China by X-ray fluorescence (XRF) and inductively coupled plasma mass spectrometry (ICP-MS). The precision of the XRF analyses is within ±2% for the oxides greater than 0.5 wt.% and within ±5% for the oxides greater than 0.1 wt. %. Analytical results for USGS standards indicated that the data are accurate within ±5% for the trace elements.

Zircon U-Pb geochronology
Zircon crystals were extracted from samples by conventional heavy liquid and magnetic separation technique and then were hand-picked under a binocular microscope. The zircon grains were mounted in epoxy and polished so that the zircon interiors were exposed. Optical photographs and cathodoluminescence (CL) images were prepared to investigate internal texture and origin of the zircons to select optimal sites for analysis.
Zircon U-Pb dating using the SHRIMP secondary ionization mass spectrometry instrument was carried out at the Beijing SHRIMP Center, Institute of Geology, Chinese Academy of Geosciences, following the standard procedures described by Jian et al. (2012). The spot size of the ion beam was about 20 μm and analytical data for individual spots are the mean values of five consecutive analyses. The analytical data were processed using the software programmes SQUID 1.03 (Ludwig 2001) and ISOPLOT 3.0 (Ludwig 2003). Errors of individual analyses are given at the 2-sigma level and the ages reported in the paper are weighted mean 206 Pb/ 238 U ages with common Pb correction using the 204 Pb-based methods of Compston et al. (1984).
Zircon U-Pb dating by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) was conducted at the Beijing GeoAnalysis Co., LTD. Pre-ablation was conducted for each spot analysis using 5 laser shots to remove potential surface contamination. The analysis was performed using 30 μm diameter spot. The Iolite software package was used for data reduction (Paton et al. 2010). Zircon 91500 was used as primary reference material, whereas zircon GJ-1 (609 Ma; Jackson et al. 2004) and Plešovice (337 Ma; Sláma et al. 2008) were used to check accuracy. Typically, 35-40 seconds of the sample signals were acquired after 20 seconds gas background measurement. The exponential function was used to calibrate the downhole fractionation (Paton et al. 2010). The analyses yielded weighted mean 206 Pb/ 238 U ages of 605 ± 4 Ma for GJ-1 and 338 ± 2 Ma for Plešovice, which agree with the reference values.

Geochemistry
Major and trace element compositions of the analysed samples are presented in Supplementary Table DR1.

Geochronology
Four samples were selected for SHRIMP or LA-ICP-MS U-Pb zircon dating. The analytical data are given in Supplementary Table DR2  DR1 b). Twenty-six zircon grains were analysed, and the measured U and Th concentrations vary from 101 to 2251 ppm and 25 to 5081 ppm, respectively, yielding Th/U ratios of 0.10 to 2.26, indicating a typically magmatic genesis for the zircon crystals (Corfu et al. 2003). All the analyses plot on and near the concordia line and yielded a weighted mean 206 Pb/ 238 U age of 316 ± 3 Ma (MSWD = 1.70; Figure 6(b) that is interpreted as the emplacement age of the gabbroic diorite.
Zircon grains from the central part of the Bagannur pluton (sample MOE-30, 108°33′54″, 47°42′6″N) are stubby prisms with grain sizes of 150-300 μm that show typical oscillatory growth zoning in CL images (Supplementary Figure DR1 c). Twenty-one zircon grains or zircon fragments were analysed, and the measured U and Th concentrations vary from 92 to 1505 ppm and 15 to 880 ppm, respectively, with Th/U ratios ranging from 0.13 to 1.05. Twelve analyses yielded consistent results giving a weighted mean

Age of the magmatism
The Middle Gobi volcanic-plutonic belt was suggested to form during the Permian-Triassic (Zorin 1999;Badarch et al. 2002), and represent an Andean-type continental margin or transform margin of the MOO. However, only a few dating results are available for the magmatic rocks in this belt. Zhao et al. (2017) reported three Permian peraluminous granitic plutons in northeastern Mongolia (273-254 Ma). Triassic ore-bearing granites were reported from the northeastern part of this belt, with zircon U-Pb ages of 241 ± 3 Ma and 229 ± 2 Ma Liu et al. 2010). These results demonstrate the Permian-Triassic intrusive magmatism of this belt. Our new age data revealed that the Adaatsag dacite, Northern Bayanjargalan gabbro-diorite and Bagannur plutons were emplaced at ca.325 Ma, ca. 316 Ma and ca. 305 Ma, respectively. They are distributed parallel to the Mongol-Okhotsk suture to the northwest of the Middle Gobi volcanic-plutonic belt, and without doubt will shed new light on the tectonic evolution of the MOO.
The ages of trapped zircon xenocrysts within the Bagannur pluton match well with the ages of zircons recovered from the Ereendavaa terrane (Miao et al. 2017;Narantsetseg et al. 2019). The inherited zircons with age spectra of 418-427 Ma is likely sourced from the Early Devonian (ca. 418-409 Ma) bimodal volcanic suite in the Ereendavaa terrane, which was interpreted to be formed by the post-collisional break-off of the subducted slab (Narantsetseg et al. 2019). The inherited zircons with ages from 434-511 Ma and 829-2433 Ma probably originated from Early-Middle Palaeozoic subductionrelated magmatic rocks and the Precambrian crystalline basement of the Ereendavaa terrane, respectively (Miao et al. 2017;Narantsetseg et al. 2019). The results indicate that the Carboniferous magmatic belt possibly developed on the basement of the Ereendavaa terrane.

Adaatsag dacite
The Adaatsag dacite shows high SiO 2 (63.75-66.96 wt. %) and low MgO (0.56-1.44 wt. %), Mg# (0.24-0.43), Cr (<10 ppm), and Ni (<5 ppm) values, indicating that they unlikely formed through partial melting of the mantle irrespective of the amount of H 2 O and volatile contents within the mantle (e.g. Mitchell and Grove 2015). Generally, there are three possible mechanisms for the generation of silicic volcanic rocks, namely (1) magma mixing of basaltic magma with felsic magmas (Kemp et al. 2007;Gao et al. 2015); (2) extensive assimilation and fractional crystallization (AFC) processes operating on a mantle-derived basaltic parental magma. (Daly 1914;Charlier et al. 2007;Davidson et al. 2007); (3) partial melting of mafic to intermediate igneous sources (Roberts and Clemens 1993;Gillis 2008;Pallister et al. 2017). The lack of mafic xenoliths (e.g. Gagnevin et al. 2008;Wang et al. 2011) and low Mg# values identical to metabasaltic rock-derived melts (e.g. Rapp and Watson 1995) indicate the dacite was unlikely resulting from the mixing of basaltic magma with felsic magma. Furthermore, AFC processes usually result in continuous compositional trends (e.g. Wanke et al. 2019), which are not the case for the Adaatsag dacite where mafic endmembers and intermediate igneous products did not occur. Moreover, the major elements concentrations (e.g. MgO) do not correlate linearly with the SiO 2 concentrations (Figure 7(a)), which further precludes magma mixing and fractional crystallization. Thus, we propose that the dacite was likely formed from the partial melting of mafic to intermediate igneous sources. The sodic feature of the dacite also requires a derivation from a low-K basaltic source, such as oceanic crust or juvenile continental crust (Rapp and Watson 1995;Sisson et al. 2005). The dacites show apparent enrichment of LREE and seems to have undergone a weak degree of feldspar fractionation as indicated by their negative Eu anomalies. They are enriched in LILE (e.g. Rb, Ba) and highly incompatible lithophile elements (e.g. U), and depleted in Nb, Ta and Ti, indicating genesis in a subduction zone (e.g. Stern 2002;Bonin 2004). In the Rb vs. (Y+ Nb) and Y vs. Nb tectonic discrimination diagrams (Pearce et al. 1984), samples of the dacite fall into the volcanic arc field, similar to the other Permian-Triassic subduction-related magmatic rocks in Mongolia (Figure 8(a, b)). Thus, we conclude that the Adaatsag dacite was formed in a mature volcanic arc environment by partial melting of juvenile mafic arc crust.

Northern Bayanjargalan gabbro-diorite
Mantle-derived magmas are often contaminated by continental materials during magma ascent. However, the very low Th/Ce (~0.05) and Th/La (~0.10) ratios of the northern Bayanjargalan gabbro-diorite rule out significant crustal contamination in its genesis, because the continental crust has relatively high Th/Ce (~0.15, Taylor and McLennan 1995) and Th/La (~0.30, Plank 2005). The gabbro-diorite shows LREEs and LILE (U, Rb and Ba) enrichment, and depletion of Nb-Ta, indicating that they may have been derived from a mantle source metasomatized by subduction-derived components (e.g. Stern 2002;Bonin 2004;Niu 2009). This hypothesis is further supported by a series of tectono-magmatic discrimination diagrams. In the Th/Y versus Nb/Y diagram (Pearce 2008), all of the samples fall above the MORB-OIB array (Figure 8(c)), indicating they are affected by the subduction processes. In the Hf-Th-Ta diagram (Wood 1980), all the gabbro-diorite samples plot in the continent arc field, similar to the low silica adakites derived from a source modified by melts from a southward-subducted Mongol-Okhotsk slab (Figure 8 (d)) (Sheldrick et al. 2020a). The samples from the Northern Bayanjargalan gabbro-diorite have negative ΔNb values (where ΔNb = 1.74 + log (Nb/Y)−1.92 log (Zr/Y)), indicating a dominantly metasomatised subcontinental lithospheric mantle source, similar to the volcanic samples from Eastern Asia older than 107 Ma (Sheldrick et al. 2020c) (Figure 9). Hence, we infer that the Northern Bayanjargalan gabbro -diorite was formed by direct partial melting of a mantle source that might have been metasomatized by fluids or melts released from subducted slab in a magmatic arc setting.

Bagannur pluton
The granite from the Bagannur pluton shows geochemical affinities of adakites, characterized by unusually high Sr/Y and La/Yb values and low Y and HREE concentrations relative to normal arc rocks (Defant and Drummond 1990;Martin et al. 2005) (Figure 10(a,b)).
Compared with the oceanic crust-derived adakitic rocks, the Bagannur granite has higher K 2 O (K 2 O = 4.00-4.48 wt. % and K 2 O/Na 2 O = 0.83-1.01), Rb/Sr (0.22-0.50) ratios and Th (10.4-14.1 ppm) contents (Defant and Drummond 1990;Martin et al. 2005;Zhou et al. 2006;Wang et al. 2008), indicating they are unlikely originate from partial melting of subducted oceanic crust. In the second model, adakites are formed from crustal assimilation and fractional crystallization of garnet and/or amphibole from basaltic magma within the garnet stability field (Macpherson et al. 2006). If garnetbearing assemblage fraction occurred, the adakites will commonly exhibit a positive correlation between Sr/Y and Dy/Yb ratios with SiO 2 contents (Macpherson et al. 2006;Wang et al. 2008), which is not observed in the Bagannur granite samples (Figure 7(c,d)). In natural systems, hornblende fractionation is commonly accompanied by the removal of plagioclase (Moyen 2009;Ji et al. 2019). Differentiation of a hornblende -plagioclase assemblage will generate negative Eu anomalies and U-shaped REE patterns between the middle and heavy REEs in the remaining melt (Castillo et al. 1999;Macpherson et al. 2006;Wang et al. 2008). However, the Bagannur high-K adakitic lava shows a rightsloping REE pattern without apparent Eu anomalies, suggesting insignificant fractional crystallization of a hornblende-plagioclase assemblage. No clear trend between Yb/Th and P/Th, Yb and P, and Sm/Th and P/ Th further precludes significant apatite fractionation (Figure 7e, f and g). Thus, we suggest that fractional crystallization, albeit responsible for minor compositional variations, probably did not play a key role for  (Pearce et al. 1984). (b) (Y + Nb) versus Rb diagram for the Adaatsag dacites and available Permian-Triassic magmatic rocks (Pearce et al. 1984). (c) Th/Yb vs. Nb/Yb discrimination diagram for the Northern Bayanjargalan gabbro-diorite (Pearce 2008). (d) Hf-Th-Ta diagram for the Northern Bayanjargalan gabbro-diorite (Wood 1980). N-MORB-normal mid-ocean ridge basalt; E-MORB -enriched mid-ocean ridge basalt; WPB-within plate basalt; IAB -Island arc basalt; WPA-within plate alkali basalt; CAB-continental arc basalt. The geochemical data of the Permian-Triassic magmatic rocks are from Tsukada et al. 2018;Ganbat et al. 2021). The geochemical data of low silica adakites are from (Sheldrick et al. 2020a).
the genesis of the Bagannur adakites. It is also unlikely that the Bagannur adakite-like rocks formed as a result of magma mixing as they contain no mafic microgranular enclaves. In addition, the major elements concentrations (e.g. MgO) do not correlate linearly with the SiO 2 concentrations, which precludes magma mixing (Figure 7 (b)). Generally, adakites generated from the delaminated lower crust are characterized by high Ni and Cr contents due to extensive interaction with mantle peridotite during ascent (Gao et al. 2004). The Bagannur high-K adakite-like granite has low MgO (0.28-0.75 wt.%), Cr (1.6-7.5 ppm) and Ni (0.7-2.6 ppm) contents, indicating that the adakite-like magma was derived from a thickened lower crust source, rather than from a delaminated lower crust (Xu et al. 2002;Gao et al. 2004;Wang et al. 2006) (Figure 10(c, d)). There is no correlation of Cr and Ni with SiO 2 concentrations (Figure 8(c, d)), indicating the Bagannur adakites are not likely formed as a result of fractionation of the adakitic melts from the delaminated lower crust with mantle involvement. The Bagannur adakites show lower Nb/Ta ratios than the average MORB and primitive mantle, indicating they are derived from the partial melting of amphibole and hornblende eclogites (Figure 7(h)). They show different geochemical features to the reported Permian adakites in the Mongol-Okhotsk belt, which are formed from slabmelting of the subducted oceanic plate (Tsukada et al. 2018) (Figure 10).
Generally, adakitic rocks derived from a thickened lower continental crust occur either in continental collision zones (Chung et al. 2003(Chung et al. , 2005(Chung et al. , 2009 or along the   (Defant and Drummond 1990) and available Permian adakites. (c and d) Cr vs. SiO 2 and Ni vs. SiO 2 diagrams for the Bagannur pluton . The geochemical data of Permian adakites are from (Tsukada et al. 2018).
Andean-type continental margins (Garrison and Davidson 2003). According to regional geological data, there is no collision-related event in the region during the Carboniferous. To say the least, the collision between the Ereendavaa and Idermeg terranes is a possible event that might be temporally linked to the generation of the Bagannur adakite-like granite. However, occurrences of post-collisional biotite granite stocks (ca. 440 Ma) and sodic granite dikes (ca. 442 Ma) demonstrate that this collision occurred during the early Silurian . This is consistent with the occurrences of the bimodal volcanic suite and coeval A-type granitic plutons (ca. 418 Ma) in the area, which demonstrate an extensional setting during the Early Devonian (Narantsetseg et al. 2019). Thus, the Carboniferous Bagannur adakite-like granite was unlikely to result from the collisional event aforemented. Considering the elongated Bagannur pluton parallel to the Mongol-Okhotsk suture zone, we suggest that its formation was related to the southward subduction of the Mongol-Okhotsk oceanic plate beneath the Ereendavaa terrane.
Though we have provided a general argument for the petrogenesis of the Carboniferous magmatic rocks, further work are required to test these conclusions. For example, we could conduct the whole-rock isotope analysis and geochemical studies on the minerals to explore the potential of magma mixing, crustal involvement and mantle input.

Tectonic implication
In the southwestern part, the closure of the MOO is accomplished by northward subduction beneath the Siberian Craton and southward subduction beneath the Central Mongolia (e.g. Zonenshain et al. 1990;Tomurtogoo et al. 2005;Metelkin et al. 2010;Donskaya et al. 2013;Tang et al. 2016;Wang et al. 2021). Late Permian-Jurassic intrusions and coeval porphyry-type ore deposits indicate that the southward subduction of the Mongol-Okhotsk oceanic plate has begun before the Late Permian (Wu et al. 2011;Xu et al. 2013;Tang et al. 2014Tang et al. , 2016Zhao et al. 2017;Deng et al. 2019). However, the early history of southward subduction of the MOO is not well constrained, especially when the southward subduction has begun. Some researchers consider that the southward subduction under the northern margin of the Amur continent began during the Permian (Zorin 1999;. Other researchers propose that the southward subduction of the MOO has begun during the Carboniferous (Bussien et al. 2011;Sun et al. 2013). In the Erguna massif in China, the Carboniferous southward subduction of Mongol-Okhotsk Ocean was mainly inferred from the occurrences of Carboniferous granites and coeval volcanic rocks with adakitic geochemical affinities (Sun et al. 2013). However, these Carboniferous adakites are proposed to be originated from the thickened lower continental crust due to the collision of the Songnen and Erguna-Xing'an terranes Sun et al. 2013). In the Mongolian segment, the Carboniferous subduction was inferred from starting input of magmatic arc zircon grains in the sandstone samples from the Ereendavaa terrane and Middle Gobi volcanic belt (Bussien et al. 2011). However, no subduction-related Carboniferous magmatic rocks have been ever reported in this region ( Figure 11). The Adaatsag dacite, Northern Bayanjargalan gabbro-diorite and Bagannur pluton in this study delineated a Carboniferous subductionrelated magmatic belt parallel to the Mongol-Okhotsk suture zone (Figure 11), which provides robust evidence for southward subduction of MOO during the Late Carboniferous. The long duration of subduction of the MOO beneath the Siberian Craton (Devonian-Triassic; Donskaya et al. 2013) and beneath the Amur Block (Carboniferous-Jurassic; this study, Sheldrick et al. 2020a) indicates the MOO was a wide ocean, which has been proved by the palaeomagnetic data (Torsvik et al. 2012;Ren et al. 2018;Zhao et al. 2020).
Devonian magmatism is absent or insignificant in the northern margins of the Central Mongolia microcontinent, in accordance with Devonian sediments of the Ereendavaa-Middle Gobi units with no contemporaneous volcanic input (Bussien et al. 2011), and together these are indicative of a passive southern continental margin setting facing the MOO. The onset of the Central Mongolia-Tsagaan Uul continental collision in the Late Carboniferous may have triggered the beginning of an active margin along the northern coast of the Central Mongolia microcontinent (Figure 12), in accordance with the 'collision-induced subduction transference' (Zhu et al. 2011;Stern and Gerya 2018). This continental collision was evidenced by the Late Carboniferous to Early Permian bimodal volcanic rocks and peralkaline granites exposed in South Mongolia (Kovalenko et al. 2006;Yarmolyuk et al. 2008;Guy et al. 2014). The bimodal association formed during 320-290 Ma and reflects the development of rifts in South Mongolia (Kozlovsky et al. 2005;Yarmolyuk et al. 2008). The peralkaline granites in the Gobi-Tien Shan area (e.g. Khan-Bogd massif and Atas-Bogd massif) and along the Main Mongolian Lineament zone (e.g. Bum Massif, North Mandakh A-type granite) were emplaced during a narrow age range from 302 Ma to 290 Ma (Yarmolyuk et al. 2008;Blight et al. 2010). Although no consensus has been reached yet regarding their tectonic environment, most researchers believe that these formations reflect a post-orogenic extensional phase (Blight et al. 2010;Guy et al. 2014;Lamb and Badarch, 1997). In addition, the significant thrusting in the middle-late Carboniferous, a sharp decrease of arc magma in the late Carboniferous and the formation of Permian intracontinental basins in the southern Mongolia also supported the tectonic transition from subduction to post-collision (Kovalenko and Yarmolyuk 1995;Kröner et al. 2010;Zhou et al. 2021). The Carboniferous subduction-related magmatic belt identified by our current study is located on the southern side of, and close to, the Mongol-Okhotsk suture, whereas the latter Permian -Triassic pluton-volcanic belt are distributed on the southern periphery of the Carboniferous magmatic belt (Figure 11). Such a configuration of these magmatic belts indicates that the southward subduction of the MOO plate possibly changed from a steep subduction during the Carboniferous to a normal or flat subduction during  the Late Permian to Triassic, which led to the magmatic arc front propagating more than 200 km away from the hypothesized trench locality. The Olzit Late Triassic bimodal volcanic rocks together with coeval A-type granites likely represent a back-arc basin extensional environment, which may have been related to the rollback of the Mongol-Okhotsk oceanic plate during the southward subduction ).
(2) These three Carboniferous magmatic complexes were formed in a subduction -related magmatic arc setting and represent results of the southward subduction of the MOO plate, attesting that the subduction had begun during the Late Carboniferous.
(3) The synchronicity and close spatial relationship between the Carboniferous southward subduction of the MOO plate and the Central Mongolia-Tsagaan Uul continental collision indicate they are probably causally correlated.