Linking deep CO2 outgassing to cratonic destruction

Abstract Outgassing of carbon dioxide from the Earth's interior regulates the surface climate through deep time. Here we examine the role of cratonic destruction in mantle CO2 outgassing via collating and presenting new data for Paleozoic kimberlites, Mesozoic basaltic rocks and their mantle xenoliths from the eastern North China Craton (NCC), which underwent extensive destruction in the early Cretaceous. High Ca/Al and low Ti/Eu and δ26Mg are widely observed in lamprophyres and mantle xenoliths, which demonstrates that the cratonic lithospheric mantle (CLM) was pervasively metasomatized by recycled carbonates. Raman analysis of bubble-bearing melt inclusions shows that redox melting of the C-rich CLM produced carbonated silicate melts with high CO2 content. The enormous quantities of CO2 in these magmas, together with substantial CO2 degassing from the carbonated melt–CLM reaction and crustal heating, indicate that destruction of the eastern NCC resulted in rapid and extensive mantle CO2 emission, which partly contributed to the early Cretaceous greenhouse climate episode.


INTRODUCTION
Carbon exchange between the Earth's interior and exterior exerts an important influence on the surface climate through geologic time and is critical for planetary habitability. In recent years, it has been increasingly recognized that the cratonic lithospheric mantle (CLM) stores vast amounts of carbon, resulting from gradual enrichment by upward melt infiltration, in addition to the original carbon incorporated during its formation [1,2]. Carbon in the CLM can be extensively remobilized and released via continental rifting [1,3], active island arc volcanism [4,5] and plume-related magmatism [6,7], which represent three main ways proposed for mantle CO 2 emission. For example, up to 28 to 34 Mt of carbon per year (expressed as Mt C yr -1 ) may be released by continental rifting [1]. A quantitative flux estimate for the CO 2 outgassing along with the massive Tan-Lu Fault Belt in eastern China gave 70 ± 58 Mt C yr -1 [8]. Extensive CO 2 degassing (71 ± 33 Mt C yr -1 ) has been estimated through extensional faults along the entire East African Rift [3], which is even comparable to the estimates for CO 2 degassing in island arcs  Mt C yr -1 ) and mid-ocean ridges  Mt C yr -1 ) [9][10][11]. Plume-induced CO 2 outgassing has also been proposed to have the ability to have caused abrupt climate changes in Earth's history [12].
Cratons commonly retain tectonic and magmatic quiescence for billions of years [13], but some cratonic regions record extensive crustal deformation and on-craton magmatism that reflect cratonic destruction processes [14,15]. In sharp contrast to many others on Earth, the eastern part of the North China Craton (NCC) is a reactivated craton with the present-day lithosphere being made up of decoupled crust and mantle, i.e. Archean-Proterozoic crust and Phanerozoic lithospheric mantle [15]. The Archean thick (diamond-bearing) and cold lithospheric keel (>200 km) was partially or even wholly destroyed and removed, and was then replaced by a newly formed thin and hot lithospheric mantle (∼75 km), resulting in up to ∼120 km of the lithospheric keel being lost [14,16]. Reactivation of the CLM gave rise to a magmatic peak at ∼125 Ma including both mafic and felsic magmatism, C The Author(s) 2022. Published by Oxford University Press on behalf of China Science Publishing & Media Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. marking the climax of cratonic lithospheric destruction in the early Cretaceous [16,17]. Since destruction/thinning is confined to the eastern part of the NCC (its western part remains largely intact), the westward subduction of the paleo-Pacific oceanic slab underneath the eastern Asian continent in the early Cretaceous was widely advocated to have resulted in reactivation and destruction of the eastern NCC [18]. In order to examine whether cratonic lithospheric destruction results in massive mantle CO 2 outgassing, we firstly ascertain whether or not the CLM beneath the eastern NCC was initially carbon rich and whether it had been widely subjected to carbonate metasomatism, including carbon from recycled carbonates prior to destruction in the early Cretaceous. For this purpose, we analyzed magnesium (Mg) isotopes for early Cretaceous lamprophyres and collated available chemical and Mg isotopic data for mantle xenoliths in Paleozoic diamond-bearing kimberlites and early Cretaceous mafic igneous rocks as well as orogenic ultramafic massifs in the Dabie orogen located on the south margin of the eastern NCC (Fig. 1). Then, we analyzed the CO 2 components of melt inclusions (MIs) in early Cretaceous lamprophyres that can be used to calculate the CO 2 concentrations in preeruptive magmas. Finally, we considered carbonated silicate melt-CLM reaction and crustal heating as additional ways for mantle CO 2 outgassing to occur during cratonic destruction, in addition to mafic magmatism. Our results show that the CLM beneath the eastern NCC had widely interacted with carbonated melts prior to the early Cretaceous and extensive CO 2 emission had occurred as a consequence of cratonic destruction.

PRIMORDIAL CARBON IN THE CLM
The terrestrial mantle initially contained carbon resulting from accretion and core-mantle differentiation processes [11]. Recent studies provide a rigorous reconstruction of carbon concentration for the MORB source mantle and suggest that the upper mantle contains ∼30 ppm C [19]. In addition to the primordial carbon, the continental lithospheric mantle, mainly formed between 2 and 3 Ga, contains more carbon (∼89 ppm) that is incorporated into the sub-arc lithosphere via the accretion of island arcs [1]. Assuming an area of ∼1 000 000 km 2 and lithospheric mantle thickness of ∼150 km for the eastern NCC prior to thinning in the Mesozoic, the CLM beneath the eastern NCC initially contains ∼4.27 × 10 7 Mt C. In addition, the model of Foley and Fischer [1] predicts a long-term (>2 Ga) solid storage of carbon in the CLM as a result of episodic melt infiltration and redox freezing. If one considers this gradual enrichment from episodic freezing throughout the long evolution history of the ancient NCC (as old as ∼3.8 Ga; [20]), carbon concentration in the CLM beneath the eastern NCC may become much higher. Abundant diamonds were discovered in Paleozoic kimberlites from the NCC [21] and are direct evidence for a C-bearing, reduced CLM beneath the eastern NCC prior to the Mesozoic era. For example, the eruption of diamond-bearing kimberlites and the high Mg # (>90; 100 × molar Mg/(Mg + Fe 2+ )) of olivines in diamond inclusions and xenolith/xenocryst olivines from Mengyin and Fuxian in the eastern NCC ( Fig. 1) indicate the existence of a thick, low-density, cold root, which is mainly composed of refractory harzburgite and lherzolite [21].

ADDITIONAL CARBON FROM RECYCLED CARBONATES
Since the Paleozoic, the NCC further underwent multiple oceanic plate subductions from the south, north and east sides, which potentially added surface carbon into the CLM beneath it. Below we present lines of evidence for more recent carbon addition to the CLM beneath the eastern NCC, from recycled carbonates related to slab subduction.

Finding of low-δ 26 Mg lamprophyres
Magnesium isotopes are a novel and efficient tool for identifying recycled carbonates that are isotopically much lighter than the mantle [22,23]. Lamprophyres are typically characterized by a high content of volatiles and commonly record fluid/melt-mantle interaction in their magma sources [24], thereby providing an opportunity to investigate the nature of fluids/melts responsible for CLM metasomatism. Early Cretaceous lamprophyres are widely exposed in the NCC and have a magmatic peak at ∼125 Ma ( Fig. 1), which is contemporaneous with the climax of cratonic lithospheric destruction of the eastern NCC [25].
Here we present a Mg isotopic dataset for Shandong lamprophyres, and for comparison we collate chemical and Sr-Nd isotopic data for other early Cretaceous lamprophyres widely distributed in the eastern NCC ( Fig. 1; Tables S1-S3). Some High-Ti lamprophyres from Shandong have relatively depleted Sr and Nd isotopic compositions ( Fig. 2a) and were proposed to have been derived from the asthenospheric mantle [26]. Most of the early Cretaceous lamprophyres in the eastern NCC, including those reported in this study, however, have extremely enriched Sr and Nd isotopic compositions ( 87 Sr/ 86 Sr (i) = 0.70520-0.71099, ε Nd (t) = −18.8 to −8.3) that are in sharp contrast to the High-Ti lamprophyres and Cenozoic alkali basalts in the NCC (Fig. 2a), pointing to an enriched CLM source. According to MgO contents, the Shandong lamprophyres are classified into Low-MgO (MgO It has been well demonstrated that Mg isotope fractionation during mantle partial melting and magma differentiation is limited (<0.07 ) [23]. In fact, the negative correlation between δ 26 Mg and MgO (Fig. 2c) [29] are shown for comparison. The magnesium isotopic composition of the terrestrial mantle (expressed as δ 26 Mg in per mil relative to DSM3) is from ref. [23]. The results of Mg-Sr isotopic modeling show a two-stage source metasomatism: the first stage is associated with siliciclastic sediments, and the second stage is related to carbonate metasomatism. The parameters used for Mg-Sr isotopic modeling are listed in the Supplementary Data (Table S4).
content (High-MgO, 81.96 ± 80.43 ppm; Low-MgO, 154.37 ± 81.89 ppm) in comparison with Low-MgO lamprophyres (Fig. S1), because marine carbonates are commonly Al and Ni poor and carbonate metasomatism can dramatically increase CaO of the mantle [27]. They also have distinct (Ti/Eu) N ratios of 0.29 ± 0.03 (High-MgO; except for one sample with 0.53) and 0.43 ± 0.07 (Low-MgO), respectively. Mantle carbonate metasomatism also accounts well for the relatively low SiO 2 of the High-MgO lamprophyres (Fig. S1) since partial melting of carbonated peridotites generates more Si-unsaturated melts relative to melting of volatile-poor peridotites [27]. The Mg-Sr isotopic mixing model (details listed in Table S4) suggests that the light-δ 26 Mg lamprophyres require source metasomatism by at least ∼10% recycled Mg-rich carbonates (e.g. dolomites; Fig. 2e). During slab subduction, Ca-rich carbonates can be substantially dissolved by aqueous fluids at initial stages and injected into the sub-arc mantle [9]. At larger depths of >160 km, dolomite dissolution occurs in subducting slabs and can be further enhanced by supercritical fluids [28]. The lithospheric mantle of the eastern NCC had a thickness of >200 km prior to thinning in the early Cretaceous [14,15]. Thus, the finding of low-δ 26 [28,34,35,57,58] are also shown (see text for details). Experimental data for carbonate melt-peridotite interaction are from ref. [36] and references therein. The data for depleted MORB source mantle (DMM) and clinopyroxene (CPX) of Paleozoic kimberlites, Mesozoic mafic rocks and Mesozoic alkaline rocks are summarized in Table S5.
Paleozoic diamond-bearing kimberlites, Mesozoic and Cenozoic mafic igneous rocks, and ultramafic massifs in the Dabie orogen derived from the deep mantle wedge (>160 km) on the south margin of the NCC (Fig. 1), record carbonate metasomatism of the CLM beneath the NCC. Wehrlites represent rocks where all, or most, orthopyroxene has been consumed through metasomatic reactions and are considered to be one of the end products of carbonate metasomatism in the CLM [8,31]. Pyroxenites and garnet pyroxenites represent rocks where all olivine and orthopyroxene have been consumed through metasomatic reactions with SiO 2 carried by supercritical fluid or silica-rich melt and are therefore considered to be other end products of carbonate metasomatism in the CLM. Abundant wehrlite xenoliths have been found in Mesozoic basaltic rocks from Tietonggou and Liaoyuan (Fig. 1a) [32,33], which suggests pervasive carbonate metasomatism of the CLM. Pyroxenite xenoliths hosted by the Jiagou intrusion (∼130 Ma) in the southeastern NCC ( Fig. 1) have extremely low δ 26 Mg of −1.23 to −0.73 (Fig. 3a) [34]. These pyroxenite xenoliths have a metasomatic U-Pb isotopic age of ∼400 Ma, suggesting carbonate metasomatism induced by paleo-Tethys slab subduction. Garnet pyroxenites in the Maowu ultramafic massif have low δ 26 Mg of −0.99 to −0.65 and contain abundant carbonate mineral inclusions and metasomatized zircons with high δ 18 O SMOW (up to 12.2 ), suggesting metasomatism of the CLM by recycled carbonates. The age of zircons (457 ± 55 Ma) from the garnet clinopyroxenites also indicates Paleozoic metasomatism by subduction of the paleo-Tethys oceanic slab [28]. Some pyroxenite and garnet pyroxenite xenoliths hosted by Cenozoic basalts (e.g. Hannuoba) also have low δ 26 Mg [35] (Fig. 3a). Because Hannuoba is located to the west of the Daxing'anling-Taihang gravity lineament (DTGL) in the western part of the NCC, in which the CLM has not been affected by the Mesozoic thinning, the presence of low-δ 26 Mg xenoliths also indicates mantle carbonate metasomatism of the NCC prior to Cenozoic. It is noted that the Hannuoba xenoliths have δ 26 Mg (low to −1.42 ; Fig. 3a) much lower than those of the host basalts and all other Cenozoic basalts in eastern China (−0.6 to −0.3 ) [29], thus their low δ 26 Mg is unlikely to have been caused by interaction between low-δ 26 Mg basaltic melt and the overlying lithospheric mantle. Generally, there is a negative correlation between δ 26 Mg and CaO content for these xenoliths (Fig. 3b), strongly suggesting metasomatism of the CLM by recycled carbonates. Apart from Mg isotopes, Ca/Al, (La/Yb) N and Ti/Eu ratios of clinopyroxenes are effective indices of mantle carbonatitic metasomatism. As shown in Fig. 3c and d, clinopyroxenes in mantle xenoliths hosted by Paleozoic diamond-bearing kimberlites and Mesozoic mafic igneous rocks have systematically higher Ca/Al and (La/Yb) N and lower Ti/Eu ratios than those of silicate-metasomatic mantle xenoliths and depleted-MORB-mantle (DMM) peridotites. Along with high Mg # and low Ti/Eu, these xenoliths are believed to have undergone carbonatitic metasomatism [36].
Further evidence for pre-Cenozoic carbonate metasomatism of the CLM comes from carbonatites. The solidus of mantle rocks can be reduced by addition of volatiles such as CO 2 into the mantle and melting of the CO 2 -rich mantle would produce alkali-rich and silicon-poor melts, such as carbonatites [27]. Mesozoic carbonatites are exposed at more than 10 locations in the NCC [36][37][38] (Fig. 1). The enriched Sr and Nd isotopic compositions of these carbonatitic magmas suggest an enriched, carbonated mantle source [37]. Mesozoic carbonatites from Zhuolu and Huairen have high 87 Sr/ 86 Sr ratios (0.7055-0.7075) and are proposed to have formed by direct melting of recycled sedimentary carbonates in the mantle [38].  Figure 4. A cartoon representing multiple subduction events around the NCC at or prior to the early Cretaceous, including the paleo-Asian oceanic slab in the north since Ordovician, paleo-Pacific oceanic slab in the east since Mesozoic and paleo-Tethyan oceanic slab in the south beginning in the Carboniferous (modified from ref. [59]). Recycled carbonates would be reduced to diamond or graphite at depths of >150 km [40].
paleo-Asian oceanic slab beneath the NCC before the Mesozoic era [39].
Collectively, the lines of evidence above strongly suggest that the CLM beneath the eastern NCC has been subject to pervasive carbonate metasomatism since the Paleozoic. The carbonate metasomatism could have been induced by multiple oceanic plate subduction events around the NCC, that is, the paleo-Asian oceanic slab in the north, paleo-Tethys oceanic slab in the south and paleo-Pacific slab in the east (Fig. 4). These subducted slabs carried large amounts of carbonate sediments into the mantle and transformed the CLM into a vast store for carbon. However, the deep part of the mantle is commonly too reduced to favor stable carbonates. That is, when carbonates are recycled into the mantle at depths of >120 km, they will be reduced via the following redox reaction: At depths of 120-170 km, recycled carbonate is transformed into carbon that exists as graphite and at larger depths (>170 km) as diamond [40], although in the CLM diamond is stable to lower pressures at cool conductive geotherms (Fig. 4). It is difficult to quantify the flux of recycled carbon in the mantle of the entire NCC since the Paleozoic, but we can give a rough estimate for this study area. As discussed above, the Mg-Sr isotopic mixing model indicates that the mantle source of low-δ 26 Mg lamprophyres contains ∼10 wt% Mg-rich carbonates (Fig. 2e), which is roughly equivalent to ∼1 wt% C. Assuming a density of 3.2 g cm −3 and a possible 40-km lithosphere depth interval that has been metasomatized, the mass of recycled C in the CLM beneath Shandong peninsula can be calculated. The lithosphere beneath the eastern NCC was >200 km thick before destruction [18,41] and the depth at which Mg-rich carbonates start to dissolve is ∼160 km [28,29]; thus we assume an ∼40-km interval for carbonate metasomatism. An areal estimate is available for the Shandong peninsula (∼73 000 km 2 ), and we assume that about half the area was affected based on the proportion of occurrence of High-MgO lamprophyres with low δ 26 Mg in the study area (Fig. 2). From this, 6.09 × 10 7 Mt C is estimated to have been added by recycled carbonates. Together with the primordial carbon (∼4.27 × 10 7 Mt C) in the CLM prior to the Paleozoic, the total reservoir of carbon in the CLM beneath the eastern NCC would be 1.04 × 10 8 Mt C at least, which represents a significant store of carbon in the CLM with important contribution from recycled carbonates. The reduced CLM, with carbon mainly existing as graphite or diamond, has not undergone redox melting and was preserved until the Mesozoic during which it was largely activated and removed.

DEEP CO 2 OUTGASSING INDUCED BY CRATONIC DESTRUCTION
Commonly, the deep CLM is primarily reduced as a result of depletion in basaltic melt and the pressure effect on the oxygen fugacity during its formation [42]. A reduced CLM beneath the thick NCC (>200 km) is indicated by the Paleozoic diamond-bearing kimberlites. However, it can become oxidized as the diamondiferous CLM is exhumed to shallower depths due to lithospheric thinning, extension and mantle upwelling [8]. During this process, 'redox melting' would occur and carbon (diamond or graphite) in the CLM would become unstable and be oxidized by the reduction of Fe 3+ at depths of <170 km [40]. This redox melting would produce carbonatitic melts at depths of ∼150 km that could evolve into carbonated silicate melts accompanied by silicate melting at shallower depths. The redox melting of the CLM was probably induced by the decompression and rise of the lithosphere-asthenosphere boundary due to slab rollback of the westward subducting paleo-Pacific plate at the early Cretaceous [18]. Carbonatitic melts have much lower viscosity and density relative to silicate melts [43], which could further promote carbonatite metasomatism of the shallower CLM. In the presence of carbonatitic melts, the mantle could be readily fusible, leading to efficient extraction of carbon from the deep interior [11]. Therefore, given the extensive thinning (>120 km) of the lithospheric mantle keel of the eastern NCC (e) Raman spectrum of a CO 2 -bearing bubble in a clinopyroxene-hosted MI from lamprophyres. The presence of CO 2 is confirmed by the Fermi diad, consisting of two peaks at ∼1285 cm -1 and ∼1388 cm -1 , bounded by hot bands, below 1285 cm -1 and above 1388 cm -1 . (f) The relationship between the volume of CO 2 in bubbles and the CO 2 concentrations of MIs. [14,16], extensive CO 2 outgassing is expected to have occurred as the CLM underwent redox melting during this thinning process. Here we evaluate whether or not the early Cretaceous lamprophyres are CO 2 rich by analyzing gas exsolution bubble-bearing MIs in them. MIs are small droplets of silicate melt trapped by crystals in magmatic rocks and can be used to constrain the contents of volatile components dissolved in melts prior to volcanic eruption and, ideally, degassing. After the MIs are captured, bubbles will be formed during the cooling process of melts, post-entrapment crystallization on MI walls, or diffusive H + loss [44]. Thus, the CO 2 of MIs is present mainly in bubbles due to its low solubility in silicate melts if postentrapment degassing occurs [44]. MIs are mainly hosted in clinopyroxene and occasionally in olivine and amphibole macrocrysts of the Shandong lamprophyres (Fig. S3). The compositions of gas exsolution bubbles were analyzed by Raman spectroscopy (see Supplementary Data for detailed methods). Among the ∼200 MIs we analyzed, >80% of the MIs contain vapor bubbles and ∼20% of the vapor bubbles contain CO 2 . The analyzed bubbles in most MIs are composed of pure or nearly pure CO 2 without other volcanic gases (CO, CH 4 , H 2 S, H 2 O) be-ing detected. The presence of CO 2 in the bubbles of MIs was confirmed by two characteristic peaks, at ∼1285 cm -1 and ∼1388 cm -1 , defining a Fermi diad in the Raman spectrum (Fig. 5). CO 2 density (d) of the bubbles can be calculated by the spacing of the Fermi diad ( cm -1 ), using the equation of Kawakami et al. [45]. The mass of CO 2 in bubbles can be calculated by multiplying CO 2 density by the volume of the bubble (Table S6). Then, the CO 2 content of the vapor bubble in ppm, [CO 2 ] vb , can be calculated using the following equation [44]: where M gl is the mass of glass within the MI, calculated as the glass volume multiplied by a melt density that is assumed to be 2.75 g cm −3 [46]. The results show that bubbles in MIs from the lamprophyres contain 323 to 47 490 ppm CO 2 (N = 29), and >93% of bubbles have a CO 2 content of >1000 ppm (0.1 wt%) ( Fig. 5f; Table S6). The calculated CO 2 concentrations in MIs of the lamprophyres range from 474 to 47 641 ppm (N = 29), with most (>80%) higher than 5000 ppm. Silicate crystal-hosted MIs, representing melts during various stages of an evolving magmatic system, can be analyzed to constrain the CO 2 contents that dissolved in the melt before volcanic eruption and/or degassing [47]. We thus estimate that the measured CO 2 concentrations represent those of the pre-eruptive and possibly evolved lamprophyre magmas, which mostly fall between 0.5 wt% and 2.0 wt%. These contents are similar to or even higher than the CO 2 concentrations (0.5-1.0 wt%) in MIs of the end-Triassic Central Atlantic Magmatic Province basalts, which were estimated by the same method [48]. It should be noted that a high CO 2 concentration of MIs is mainly observed in High-MgO lamprophyres with light δ 26 Mg values (see Supplementary Data), and the number of MIs in High-MgO lamprophyres is much larger than that in Low-MgO lamprophyres. This probably indicates that High-MgO lamprophyres with recycled carbonates in their mantle sources contain more MIs and higher CO 2 concentrations in the pre-eruptive magmas, although low-volume melts could also have extremely high CO 2 content even if the source is not specifically C rich, due to the strong incompatibility of CO 2 in peridotite [49]. Because most of the early Cretaceous lamprophyres in the eastern NCC belong to the High-MgO group with low δ 26 Mg (Fig. 2), their sources were plausibly most strongly affected by carbonate metasomatism and attendant enrichment in carbon. Melting of this metasomatized CLM then produced primary magmas with high CO 2 content, which may have been further enhanced during pre-eruptive differentiation. Here we collated geochemical data for early Cretaceous mantle-derived volcanic rocks (see Fig. 1 for locations, Fig. 2a and b and Fig. S4 for chemical compositions) and found that they are widely distributed in the eastern NCC and show similar geochemical characteristics to the lamprophyres. Thus, early Cretaceous mantle-derived magmas in the eastern NCC are much more abundant than those represented by lamprophyres. A larger flux of CO 2 outgassing is thus expected during the period of extensive destruction of the NCC, in addition to lamprophyres. Intrusion or eruption of these magmas could have carried a large amount of CO 2 from the mantle into the surface. Experimental studies show that the solubility of carbon dioxide in melts decreases at lower pressures, and CO 2 can even be directly degassed at mantle depths [50]. Therefore, the thinned lithosphere, as a result of cratonic destruction and extension, can further facilitate CO 2 outgassing via magmatism. There is no evidence for the presence of a deep-sourced mantle plume beneath the eastern NCC during the Phanerozoic era. We thus propose that the destruction of the CLM represents another important cause of CO 2 emission from the mantle, in addition to continental rifting, active island arc volcanism and mantle plume. During this process, carbon in the CLM can experience gradual oxidation during mantle upwelling, with a change of carbon speciation from a reduced to an oxidized form, and a portion of carbon can be liberated via redox melting from the reduced mantle [11].
Enormous CO 2 reservoirs can be formed by the eruption or intrusion of magmas. There are abundant crustal CO 2 reservoirs in the Songliao and Bohai Bay basins in the eastern NCC, which indicates that the volume of CO 2 degassing is enormous [51]. CO 2 reservoirs in the Songliao basin were formed primarily in Cretaceous, and voluminous inorganic CO 2 (mainly mantle-derived and crust-derived) is observed in these reservoirs. For example, the high CO 2 content (>90%) and δ 13 C (−4.95 ) and high helium isotopic composition (R/R a = 3.34) of Wanjinta reservoirs indicate that the CO 2 was chiefly sourced from the mantle [51]. The Bohai Bay basin, a Mesozoic-Cenozoic basin, is the central area of destruction of the eastern NCC. The reservoirs there also have high CO 2 content (79.17-98.61%) and high R/R a (2-3.34), which indicates that the CO 2 was derived from the mantle [51].
A recent study by Aulbach et al. [8] quantified the CO 2 flux related to the reaction of the CLM with silica-undersaturated (carbonated) melt, referred to as wehrlitization, and linked this flux to surficial de-gassing in rifts and basins. As discussed above, abundant wehrlite and pyroxenite xenoliths are found in Mesozoic mantle-derived rocks in the eastern NCC, and many of these xenoliths have light Mg isotopic compositions and high Ca/Al ratios (Fig. 3). For instance, the characteristics of low Ti/Eu, high Ca/Al, (La/Yb) N and Zr/Hf of clinopyroxenes in Liaoyuan wehrlites are ascribed to interaction with a silicaundersaturated, carbonated silicate melt [33]. These rocks thus record the substantial reaction between carbonated silicate melts and the CLM. In the CLM, these melts are initially out of thermal and compositional equilibrium, causing intensive melt-rock reactions. During this process, the following reaction will happen at ∼1.5-2.0 GPa [52]: A quantitative estimate suggests that 2.9 to 10.2 kg CO 2 can be released per 100 kg of wehrlite formed [8]. Extensive CO 2 release is suggested to have occurred during the carbonated melt-CLM reaction process in the course of this destruction of the eastern NCC, along with the Tan-Lu Fault Belt, which was most active in the early Cretaceous [33].
Stable continents are long-term storage sites for sedimentary carbonates, and the amount of carbonates stored in continents is thought to be at least 10 times greater than that stored in oceanic crust [53]. Carbonates in crusts can be trapped by plutons that ascend to shallow levels in the arc crust or are transported into the lower crust during later arc stages. Global flare-ups in continental arc volcanism were proposed to have the potential to release CO 2 as a result of magmatic interaction with ancient crustal carbonates stored in the continental crust [54,55]. The eastern NCC is typically characterized by a giant felsic magmatism event at the early Cretaceous with a volume much larger than that of mafic magmatism [17], implying large-scale crustal melting and reworking during the cratonic destruction process. These early Cretaceous felsic magmas (i.e. granites) have high zirconium saturation temperatures and contain an important contribution from the hot upwelling mantle [25]. Decarbonation is expected to widely occur during interaction between the hot felsic magmas and the limestones chronically stored in the continental crust. This process could also contribute to CO 2 release, in addition to mantle CO 2 outgassing via mafic magmatism and carbonated melt-CLM reaction during destruction of the eastern NCC (Fig. 6). Rifting diffusing Cratonic destruction Figure 6. Schematic cartoon models illustrating mantle CO 2 outgassing in response to cratonic destruction. The carbonated CLM beneath the eastern NCC underwent extensive melting and thinning at ∼125 Ma as a result of heat upwelling from mantle convection related to rollback of the subducting west Pacific slab. This process produced extensive CO 2 -rich magmas with a peak at ∼125 Ma, as recorded by lamprophyres, basalts and carbonatites. The CO 2 vapors were released in three ways: 1 magmatic degassing of lamprophyre and eruptive magmas, 2 heating of sedimentary carbonates stored in the crust, represented by granites, and 3 the interaction between carbonated melts and the CLM, represented by wehrlites and pyroxenites leading to decarbonation and liberation of CO 2 vapor.

POSSIBLE CONTRIBUTIONS TO THE CRETACEOUS GREENHOUSE
The amount of CO 2 outgassing induced by the destruction of the eastern NCC could be significant, particularly if one considers that the removed CLM contained a large amount of recycled carbon from subducted slabs prior to thinning. The fluxes of deep CO 2 outgassing during this destruction process may be large given the short duration of mantlederived, CO 2 -rich magmatism (both intrusions and volcanics; Fig. 1) in the early Cretaceous (∼125 Ma) as well as the strong carbonated silicate melt-CLM reaction that resulted in substantial CO 2 release along the massive Tan-Lu Fault Belt [8]. At a larger scale, enormous quantities of CO 2 that were rapidly released into the atmosphere, induced by the destruction, may have perturbed the global climate and partly contributed to the atmospheric CO 2 rise during the Cretaceous, one of the longest greenhouse periods of Earth's history, with atmospheric CO 2 levels 4 to 10 times higher than those prior to the Industrial Revolution [56].

CONCLUSIONS
We present the first Mg isotope data for early Cretaceous lamprophyres and collate available chemical and Mg isotopic data for mantle xenoliths in Paleozoic diamond-bearing kimberlites and Mesozoic mafic igneous rocks as well as orogenic ultramafic massifs in the NCC. These results suggest the presence of a widespread, C-rich CLM beneath the NCC at and before lithospheric thinning in the early Cretaceous, with an important contribution from recycled carbonate sediments. Long-term and threesided-i.e. the south, north and east-oceanic plate subductions underneath the NCC during the Paleozoic and Mesozoic could have contributed a vast amount of carbon to the lithospheric mantle of the NCC. Redox melting of the reduced, C-rich CLM as it was exhumed to shallower depths due to lithospheric extension and thinning in the early Cretaceous generated large amounts of basaltic lavas and lamprophyres, resulting in the release of voluminous CO 2 to the exosphere. This may represent an important cause of CO 2 emission from the mantle, in addition to mantle plumes, active island arc volcanism and continental rifts, as proposed in previous studies [1,[3][4][5][6][7]. The amount of magmatic CO 2 outgassing is largely supplemented by the release of mantle CO 2 induced by the carbonated melt-CLM reaction [8] and decarbonation induced by interaction between hot felsic magmas and crustal limestones. Therefore, deep CO 2 outgassing can be linked to the destruction of a long-term stable craton and can be said to have enhanced global CO 2 input into the atmosphere.