Origin of unusual HREE-Mo-rich carbonatites in the Qinling orogen, China

Carbonatites, usually occurring within intra-continental rift-related settings, have strong light rare earth element (LREE) enrichment; they rarely contain economic heavy REE (HREE). Here, we report the identification of Late Triassic HREE-Mo-rich carbonatites in the northernmost Qinling orogen. The rocks contain abundant primary HREE minerals and molybdenite. Calcite-hosted fluid inclusions, inferred to represent a magmatic-derived aqueous fluid phase, contain significant concentrations of Mo (~17 ppm), reinforcing the inference that these carbonatitic magmas had high Mo concentrations. By contrast, Late Triassic carbonatites in southernmost Qinling have economic LREE concentrations, but are depleted in HREE and Mo. Both of these carbonatite types have low δ26Mg values (−1.89 to −1.07‰), similar to sedimentary carbonates, suggesting a recycled sediment contribution for REE enrichment in their mantle sources. We propose that the carbonatites in the Qinling orogen were formed, at least in part, by the melting of a subducted carbonate-bearing slab, and that 10 Ma younger carbonatite magmas in the northernmost Qinling metasomatized the thickened eclogitic lower crust to produce high levels of HREE and Mo.

melt during immiscibility (D La/Lu carbonate/silicate melts = 1.3-4.5) 15,16 . Crystal fractionation of dense silicates 17 and oxides (such as REE-enriched perovskite and pyrochlore) from carbonatites could decrease the HREE and Si contents of the residual magmas, depending on the mineral composition. Therefore, none of the three petrogenetic models noted above is likely to be responsible for significant HREE enrichment in carbonatites.
Here we present observations on primary HREE minerals and Mo-containing fluid inclusions in the NQ carbonatites, and Mg isotope systematic, obtained in order to constrain carbonatite origin. The Qinling orogenic belt separates the North China Block from the South China Block (Fig. 1a). Closure of the Mianlue Ocean in the Triassic resulted in collision of the South China Block with the Qinling belt along the Mianlue suture, followed by amalgamation of the South and North China Blocks 18 . The NQ is a world-famous province for Mo ore production: six giant Mo deposits and numerous large, medium and small Mo deposits are present in this region 19 . The majority of the Mo deposits formed during the Late Jurassic to Early Cretaceous (148-112 Ma), and are closely associated with granitic porphyries 19 . The NQ carbonatite samples studied in this investigation were collected from Huayangchuan and Huanglongpu (Fig. 1b). The carbonatites occur as dykes and lenses, extending from tens of meters to over one kilometer, intruding into Late Archean gneiss and Mesoproterozoic andesitic rocks 8 . The carbonatites contain abundant molybdenite and form a large Mo deposit 7 .
Primary HREE minerals were found in the carbonatites (Fig. 2) and have been grouped into xenotime-(Y) and Y-, Y(Ce)-Ca-and Y-Fe-silicate minerals in terms of their main element contents measured by EPMA (Supplementary Table S1). Xenotime-(Y) (Y 0.97-0.99 PO 4 ) occurs as irregular to subhedral crystal intergrowths with molybdenite and monazite-(Ce) (Fig. 2a,b), or allanite-(Ce) (Fig. 2e). The Y-silicate mineral [(Y,Ce,Ca) 1.9-2.1 Si 2-2.1 O 7 ·H 2 O] is potentially a hydrated form of keiviite-(Y), and forms small and parallel plate-like aggregates together with molybdenite and monazite-(Ce) (Fig. 2c,d). It is high in Y 2 O 3 (~30 wt.%) and contains minor amounts of LREE (Ce 2 O 3 = 3-6 wt.%). Minor CaO (0.2-3 wt.%) was also detected. This mineral shows a relatively constant REE distribution pattern on chondrite normalized (CN) plots and a low La/Yb CN ratio of 0.39 (Fig. 3). The Y(Ce)-Ca-silicate minerals are potentially forms of britholite-(Y) and kainosite-(Ce) 7 ], and generally occurs as laminar crystals along the margins of the quartz. Small augite crystals were also observed in the quartz (Fig. 2f,g). The HREE-silicate mineral has variable contents of Y 2 O 3 (5-22 wt.%) and CaO (4-14 wt.%) and elevated levels of LREE (Ce 2 O 3 up to 18 wt.%). It contains minor amounts of F (up to 2.6 wt.%) and shows a variable REE pattern  Table S2). These ages are close to the Re-Os age (ca. 221 Ma) 22 of molybdenite in the carbonatites, and they are different to those of the granite porphyries and associated Mo deposits, which formed during the Late Jurassic to Early Cretaceous in the NQ orogen 19 . These results suggest that the Mo enrichment is a primary feature of the carbonatite magmas, as is also supported by the fluid inclusion results. Calcite, anhydrite, mirabilite, celestine and burbankite were determined as daughter minerals in the primary fluid inclusions in the calcite and quartz, this being consistent with the trapping of a carbonatite-exsolved fluid at subsolidus conditions (Supplementary Figs S3 and S4). Small but significant amounts of Mo (16 and 17 ppm; Supplementary Table S3) were determined in the calcite-hosted fluid inclusions by in-situ laser ablation ICPMS analysis. This concentration of Mo is comparable to that of fluid inclusions within the mineralized system from the super-large porphyry Mo deposit at Questa (6 to 90 ppm) 23 . To our knowledge this is the first case of detection of Mo in carbonatitic fluids.
Besides molybdenite and HREE minerals, the presence of abundant quartz in the NQ carbonatites distinguishes them from the majority of carbonatites worldwide. The origin of some quartz-bearing carbonatites has been postulated as representing calc-silicate skarn (pseudo-carbonatite) 24,25 . However, our chronology, mineralogy and geochemistry do not support such a pseudo-carbonatite origin for the rocks. There are no sediments and contemporaneous silicate rocks spatially associated with the carbonatites. Most of the granitic porphyries in the NQ formed during the Late Jurassic to Early Cretaceous (148-112 Ma) 19 . Burbankite, generally a diagnostic mineral of the carbonatites 20 , was identified in the studied rocks. Reaction between silicate magmas and Ca-carbonate would form wollastonite as experiments suggest 24 and as is also seen in many skarn deposits 26 Table S4). They show a strong LREE and HREE fractionation pattern (La/Yb CN = 1.3-8.4) and a negligible Eu anomaly. These results are similar to those from other carbonatites worldwide [27][28][29] , but different to apatites in granitoids from various geographically occurrences 29 . Similarly, high Sr contents were determined in the carbonatitic calcites 30 , consistent with carbonates of a carbonatitic origin 27,28 .
The NQ carbonatites have C and O isotope compositions similar to normal carbonatites 7 , and they show Sr-Nd isotope ratios typical of enriched mantle sources 31 . Sulfur isotope data further support a mantle origin for the rocks (Supplementary Table S5). Molybdenite, galena and pyrite from the carbonatites display δ 34 S values varying from − 6.7 to − 7.7‰, − 8.9 to − 10.5‰ and − 6.5 to − 7.2‰, respectively. They are distinct from barite (4.6 to 5.1‰). As sulfur isotopes can be re-distributed between oxidized and reduced species during precipitation 32 , the actual δ 34 S (δ 34 S ΣS ) of the carbonatitic liquids cannot be represented by the mean δ 34 S values of either sulfides or sulfates in this case. The δ 34 S ΣS of the carbonatitic liquids has been calculated using galena and co-genetic barite as a mineral pair and using the δ -δ diagram 33 (Fig. 4). As estimated from the intersection of a straight line fit to the data with the line of unit slope which shows no sulfide-sulfate fractionation (Δ = 0‰), the δ 34 S ΣS value of the precipitating liquids is estimated to be around 1‰, similar to the mantle-derived value 34 .
Mg isotopic compositions. Carbon isotopes are commonly used to trace subducting sediments 35 .
Although they are effective in identifying recycled organic carbon, they are not sensitive to inorganic carbon which dominates subducted materials 36 . Recently, due to the fact that subducting lithologies underwent little Mg isotope fractionation during metamorphic dehydration and magmatic processes 37,38 , Mg isotope systematics have been applied to trace recycled sediments. The Mg isotopic compositions of carbonatites from the NQ, and  Table 1. The carbonatites in the southernmost Qinling host a large LREE deposit 39 . Both carbonatites have significantly lower Mg isotope ratios than typical mantle derived rocks, including Mid Ocean Ridge Basalt (MORB), Ocean Island Basalt (OIB) and peridotite xenoliths (− 0.25 ± 0.07‰) 37 , as well as intra-continental rift-hosted carbonatites from Oldoinyo Lengai (0.13-0.37‰) 40 , with δ 26 Mg ranging from − 1.28 to − 1.07‰ and − 1.89 to − 1.37‰ at NQ and Miaoya, respectively (Fig. 5).
Magnesium isotopic fractionation can result from a range of processes, including mantle melting, crystal fractionation, silicate-carbonate liquid immiscibility and crustal contamination. There is insignificant Mg isotopic fractionation during partial melting of the mantle, or from olivine and pyroxene crystallization 37,40,41 . Li et al. 40 suggested that the light Mg isotope preferentially enters into crystallizing carbonates, leading to differentiated carbonatite enriched in the heavy Mg isotope. This was identified in natrocarbonatites at Oldoinyo Lengai which have δ 26 Mg compositions of 0.13-0.37‰ 40 . Although the heavy Mg isotope has been demonstrated to be preferentially incorporated into silicate relative to carbonate melts during liquid immiscibility, such fractionation is relatively small (− 0.3 to 0.09‰) 40 . Due to the high δ 26 Mg composition of the lower, middle (− 0.26 to − 0.21‰) 42 and upper continental crust (− 0.52 to 0.92) 43 , crustal contamination will increase the Mg isotope ratio in mantle-derived magmas. The majority of magmatic fractionation processes therefore cannot produce the low levels of δ 26 Mg identified in the carbonatites in this investigation. In contrast, marine dolomite has low δ 26 Mg values of − 2.5 to − 1‰ over geological time 44 , and its Mg isotopic composition can be preserved during subduction 45 . Therefore, the carbonatite mantle sources in the Qinling orogen may contain recycled sediments.

Discussion
Several studies indicate that primary mantle-derived carbonatitic melts contain relatively low REE contents 13 50,51 . Geochemical evidence also shows that the Tromsø Nappe carbonatites in the Caledonian orogenic belt have an genetic relationship with associated eclogite 52,53 . These results imply that recycled sediments are a potential REE source for carbonatite magmas. However, the NQ carbonatites are also enriched in HREE and Mo, thus distinguishing them from the other orogenic carbonatites with LREE enrichment and Mo depletion 39,48 . Experimental studies suggest that REE and Mo do not preferentially partition into carbonatite-exsolved fluids 54 , and geochronological data are not consistent with their derivation from later igneous sources within the Qinling region. Therefore, post-magmatic fluid metasomatism cannot explain HREE and Mo enrichment in the NQ carbonatites. This indicates that the primary carbonatitic magmas may have contained high HREE and Mo concentrations, and have undergone a different evolution in the orogen. The Late Triassic carbonatites from Miaoya in the southernmost Qinling orogen are compared with the NQ rocks in Fig. 1b. The Miaoya carbonatites occur as stocks close to the collisional suture between the South and North China Blocks, and they are inferred to have formed from melting of subducted oceanic crust 18 . The rocks contain economic LREE ores and lack HREE minerals and molybdenite. They have a slightly older monazite age of 233 Ma 18 compared to the NQ monazite (209-214 Ma) and molybdenite (221 Ma) ages. Although the NQ and Miaoya carbonatites have similar light Mg isotope compositions, they show obvious differences in their initial Sr-Nd isotope systematics. The Miaoya carbonatites (initial 87 Sr/ 86 Sr = 0.7036-0.7040 and ε Nd = 0.6 to −1.1) 18 fall into the HIMU-EM1 trend, corresponding to most young carbonatites worldwide 55 . By contrast, the NQ carbonatites have more radiogenic Sr (initial 87 Sr/ 86 Sr = 0.705-0.706) and lower Nd isotopes ratios (ε Nd = − 10.1 to − 4.3) 31 . These differences can hardly be attributed to mantle heterogeneity within a relatively confined area.
The lower crust of the north part of the Qinling has higher Mo content (2.04 ppm) than that in the South Qinling, other major tectonic units in eastern China (0.52-0.55 ppm) 56 and average lower crust worldwide (0.6 ppm) 57 , and is an important Mo source. On this basis we propose the following model for the origin of HREE-Mo rich carbonatites in such an orogenic setting. The formation of both the Late Triassic NQ and Miaoya carbonatites followed the closure of the Mianlue ocean and collision of the South and North China Blocks 18 . Subducted carbonate-bearing ocean crust melted to form carbonatite magmas which intruded in a post-collisional setting. The NQ carbonatites were then emplaced about 10 Ma later than the Miaoya rocks, and they underwent a relatively longer residence time in the mantle source. There are two possible mechanisms for the NQ carbonatite magma interaction with lower crustal materials. During the inter-continental collision, the lower continental crust thickened to 37-82 km 58 , resulting in transformation of the mafic underplate to eclogite. The P-T conditions at the depth of eclogitic crust could then stabilize the carbonate melt (> 2 GPa) 9,12 . The thickened lower crust may have been directly metasomatized by the upwelling carbonate melt. Alternatively, the eclogitized crust had a higher density than the sub-continental lithospheric mantle which resulted in the delamination of lower crust into the carbonatite mantle source. Breakdown of HREE-enriched eclogitic garnets would then release the HREE into the carbonatite melts. A similar reaction was proposed based on the observation of garnet in peridotite xenoliths 59 , which shows that the metasomatized garnets have lower HREE-contents than the unmetasomatized ones.
The lower crust, when compared to mantle assemblages, is relatively rich in silicon and reactions therefore would have increased silica contents in the reacted carbonatitic melts. The presence of primary HREE-silicate minerals in the NQ carbonatites implies that silicon may play a role in HREE enrichment. Experimental results suggest that increased Si content in carbonatitic melts favors REE incorporation into apatite, particularly HREE (e.g., D La apatite/melt = 1.  61 , whereas all REE are incompatible with respect to apatite in a silica-poor system (D REE apatite/melt < 1) 62 . A study on an REE-rich syenite in Quebec, Canada, showed that HREE will preferentially partition into an immiscible Fe-Si-rich magma relative to the coeval syenite magma 63 .
In conclusion, the HREE-Mo-rich carbonatites in the NQ can be inferred to represent mixtures between carbonatite melt and a component from an eclogitic lower crust. High Mo concentrations have been determined in the lower crust from our study area 56 . Eclogite xenoliths in Mesozoic adakitic porphyries from the southeastern margin of the North China Block 64 document the existence of an eclogitic crust within the deep lithosphere (> 1.5 GPa) in the NQ. The eclogite yields a Sm-Nd whole-rock-garnet isochron and zircon U-Pb dates of 220 Ma 64 . Carbonatites associated with eclogite have also been observed in the orogenic belt of the northern Scandinavian Caledonides 52,53 . Although it is accepted that carbonatite metasomatism plays an important role in the modification of the chemistry of the lithospheric mantle 65 , it is not yet known if carbonatite melt generation involves metasomatism of thickened lower crust in orogenic belts. This process may have had the potential to result in HREE enrichment in the carbonatite magma. Apart from the Huayangchuan and Huanglongpu occurrences in this study, another Late Triassic carbonatite (molybdenite Re-Os age of 210 Ma) from Huangshuian 66 in the middle NQ ( Fig. 1) has been recently discovered, thus increasing the geographic extent of these geochemically anomalous carbonatites. Thus, overall, the NQ orogen has promising potential as a REE-mineralized belt.

Methods
Monazite U-Th-Pb dating. U-Th-Pb analyses of the monazite were performed using the Cameca IMS-1280 (SIMS) at the Institute of Geology and Geophysics, Chinese Academy of Sciences (CAS). The O 2− primary ion beam was accelerated at 13 kV with an intensity of ca. 2-3 nA. The aperture illumination mode (Kohler illumination) was used with a 200 μ m primary beam mass filter (PBMF) aperture to produce even sputtering over the entire analyzed area. The ellipsoidal spot is about 20 × 30 μ m in size. Positive secondary ions were extracted with a 10 kV potential. Sample charging effects were minimized by optimizing the energy offset to maximum transmission in a 60 eV energy window at the start of each analysis, using Th + as reference peak. A mass resolution of ca. 5,500 (defined at 1% peak height) was used, which is enough to separate U, Th, and Pb isotopes from isobaric interferences, such as REE oxides, except for 204 Pb which could be interfered by ThNdO 2 + . Monazite 44069 was used as a standard.
EPMA and LA-ICPMS analysis of minerals. The major element compositions of the HREE minerals in the carbonatites were determined using an EPMA-JXA-8230 electron microprobe with wavelength-dispersive X-ray spectrometry (WDS) at the Key Laboratory of Metallogeny and Mineral Assessment, Chinese Academy of Geological Sciences. The measurement conditions were 15 kV accelerating voltage and 20 nA beam current, with an electron beam of 2-5 μ m diameter. According to the components of separated minerals, their own set of appropriate matrix-specific standards (both synthetic and natural) and optimal instrumental conditions (detector type, counting statistics and beam settings) were carefully chosen by performing multiple measurements. All data were corrected with standard ZAF correction procedures. The raw WDS data of REE were corrected on the basis of empirical interference values for REE using well-characterized REE glasses, synthetic standards and orthophosphates.
In-situ laser-ablation (LA) ICP-MS analyses of fluorapatites were performed using an Agilent 7500Ce mass-spectrometer coupled to a 193-nm ArF excimer laser at the Peking University. The diameter of the ablation spot used was 30 μ m. NIST 610 glass was used as a calibration standard for all samples. The element used for the internal standard was Ca, measured as 43 Ca and expressed as CaO, which was independently measured by electron microprobe. The analytical error is within 5%.

Fluid inclusion measurement. Laser Raman spectroscopic analyses of compositions of the individ-
ual fluid inclusions in the carbonatites were performed using the LabRam HR800 at the Institute of Geology and Geophysics, CAS. The laser source was an argon ion laser with wavelength of 532 nm and output power of 44 mV. Count time for each measured spectrum is 3 s and counted once for each cm −1 . The spectral resolution was 0.65 cm −1 with a beam size of 1 μ m. The peak was taken from 100 to 4000 cm −1 once in full-wave band. Instrumental settings were kept constant during all the analyses. Solid minerals in opened inclusions were identified by LEO-1450vp scanning electron microscope with an attached Oxford INCA-ENERGY 300 X-ray spectrometer operated at 15 kV at the same Institute.
The composition of fluid inclusions in the carbonatitic calcite was measured by LA-ICPMS at the Advanced Analytical Centre, James Cook University, Australia. A GeoLas 193 nm ArF excimer laser coupled with a Varian 820 quadrupole ICP mass spectrometer with He as the carrier gas in the ablation chamber was used. To obtain maximum signal sensitivities, a Y-piece was used to admix Ar prior to introduction to the ICP. The laser beam diameter ranged between 24 and 44 μ m, depending on the inclusion size and depth within the sample. Chlorine and Br concentrations were quantified using the scapolite BB1 standard 67 , whereas all other elements were quantified using NIST 610 glass. Sodium concentrations, derived from microthermometry, were used as the internal standard. The SILLS software was used to perform all data reduction procedures.
Scientific RepoRts | 6:37377 | DOI: 10.1038/srep37377 Sulfur isotope analysis. Sulfur isotopic compositions of barite and sulfides from the carbonatites were measured at the Institute of Geochemistry, CAS, using a continuous-flow isotope ratio mass spectrometer (IsoPrime). Reproducibilities are ± 0.2‰, and the results are expressed conventionally as per mil (‰) variation relative to V-CDT.
Mg isotope analysis. The Mg isotopic pretreatment has been described by ref. 68. The total procedural blank for Mg was < 15 ng. The isotope was analysed using a Neptune MC-ICP-MS at the China University of Geosciences (Beijing). The analytical results are expressed as δ -notation in per mil relative to DSM3: δ X Mg = [( X Mg/ 24 Mg) sample /( X Mg/ 24 Mg) DSM3 − 1] × 1000, where X refers to mass 25 or 26 and DSM3 is a magnesium solution made from pure Mg metal. The analyses of standard BCR-2, BHVO-2, Kilbourne Hole olivine, and Hawaii seawater were in good agreement with recommended values within 2σ error 69 .