In Situ Geochemical and Sr–Nd Isotope Analyses of Apatite from the Shaxiongdong Alkaline–Carbonatite Complex (South Qinling, China): Implications for Magma Evolution and Mantle Source

: We present in situ major element, trace element, and Sr–Nd isotope data of apatite from an alkaline–carbonatite intrusion in the South Qinling Belt (SQB) to investigate their magma evolution and mantle sources. The Shaxiongdong (SXD) complex consists predominantly of the early Paleozoic hornblendite, nepheline syenite, and subordinate Triassic carbonatite. Apatites from all lithologies are euhedral to subhedral and belong to ﬂuorapatite. Elemental substitution varies from REE 3+ + Na + + Sr 2+ ↔ 3Ca 2+ in carbonatite and syenite apatite to Si 4+ + 2Na + + 2S 6+ + 4REE 3+ ↔ 4P 5+ + 5Ca 2+ in hornblendite apatite. Apatites are characterized by enriched rare earth elements (REEs) and depleted high ﬁeld strength elements (HFSEs). They record the distinct evolution of their parental magmas. The weak, negative Eu anomaly in hornblendite apatite, together with the lack of Eu anomalies in the bulk rocks, indicates a relatively reduced magma. The Sr–Nd isotope data of the apatite in SXD carbonatite, falling on the East African carbonatite line (EACL) and close to the ﬁeld of Oldoinyo Lengai carbonatite, indicate that the SXD carbonatite is derived from a mixed mantle source consisting of the HIMU component and subducted sedimentary carbonates. The similarity in Sr and Nd isotopic compositions between the SXD hornblendite and syenite apatites and the early Paleozoic maﬁc-ultramaﬁc dykes in the SQB suggests that they may share a common metasomatized lithospheric mantle source.


Introduction
Apatite, with a general formula of Ca 5 (PO 4 )(OH, F, Cl), is widespread in various environments ranging from the Earth's surface to the lithospheric mantle [1]. It is a ubiquitous accessory mineral in most igneous, metamorphic, and sedimentary rocks [2]. Apatite can also become a rock-forming mineral in phosphorites and iron oxide-apatite deposits [3][4][5]. As the most abundant phosphate on Earth, apatite is a wonderful mineral to use to explore the global phosphorus cycle [6,7]. Also, apatite has been found in some lunar basalts, indicating that the volatile phase may play an essential role in the evolution of lunar magmatism [8][9][10].
Apatite generally has variable trace element contents due to numerous substitutions. It is classified as fluorapatite, chlorapatite and hydroxyapatite based on the different column anions (F, Cl, OH). Fluorapatite and hydroxyapatite are the most commonly occurring phosphorus-bearing minerals in igneous rocks. Numerous studies have shown that apatite has a strong affinity for the REEs due to extensive substitution of REEs [11][12][13][14]. There are two major substitution mechanisms at Ca and P sites: (1) REE 3+ and Na + substitute for Ca 2+ and (2) REE 3+ and Si 4+ substitute for Ca 2+ and P 5+ [15][16][17]. As a major REE-bearing mineral the final suture between the NCB and the QOB. The QOB comprises the South Qinling Belt (SQB) and the North Qinling Belt (NQB), which were welded together along the Shangdan suture. The NQB is thought to be a part of the southern margin of the NCB, whereas the SQB is proposed to be the passive continental margin of the SCB before the opening of the Mianlue Ocean during the late early Paleozoic [43]. Early Paleozoic bimodal volcanic rocks indicate the initial spreading of the Mianlue Ocean. In the Triassic, northward subduction of the Mianlue ocean ultimately resulted in an amalgamation of the SQB and SCB [44]. A suite of mantle-derived mafic sills associated with the post-collision stage occurs in South Qinling [46].
Qinling Belt (SQB) and the North Qinling Belt (NQB), which were welded together alon the Shangdan suture. The NQB is thought to be a part of the southern margin of the NC whereas the SQB is proposed to be the passive continental margin of the SCB before th opening of the Mianlue Ocean during the late early Paleozoic [43]. Early Paleozoic b modal volcanic rocks indicate the initial spreading of the Mianlue Ocean. In the Triass northward subduction of the Mianlue ocean ultimately resulted in an amalgamation the SQB and SCB [44]. A suite of mantle-derived mafic sills associated with the post-col sion stage occurs in South Qinling [46]. The SQB is characterized by thin-skinned south-vergent imbricated thrust-fold sy tems. Two different Precambrian basements, namely the crystalline and transitional bas ments, have been distinguished in the SQB based on the lithology and metamorphic grad (Figure1b). The crystalline basement is represented by amphibolite-and granulite-faci assemblages of the Douling complex, whereas the transitional basement comprises th Meso-Proterozoic Yaolinghe and Wudang Groups of low-grade metamorphic volcan sedimentary rock assemblages [45]. In addition, Neoproterozoic tillite and platform ca bonate, Cambrian-Ordovician limestone, and Silurian shale unconformably overlie th basements.  [47]); (c) geological map of the SXD complex with the distribution of different lithologies.
The SXD alkaline-carbonatite complex, located around 12 km southeast of Zhusha County, Western Hubei Province, is tectonically situated at the southern margin of th Wudang terrane. It occurs as a spindle-like stock intruding on the Wudang Neoproter zoic metamorphic schist (Figure 1c). The complex is predominantly composed of the hor blendite outcropping in the northwestern and eastern parts of the intrusion and syeni The SQB is characterized by thin-skinned south-vergent imbricated thrust-fold systems. Two different Precambrian basements, namely the crystalline and transitional basements, have been distinguished in the SQB based on the lithology and metamorphic grade ( Figure 1b). The crystalline basement is represented by amphibolite and granulite-facies assemblages of the Douling complex, whereas the transitional basement comprises the Meso-Proterozoic Yaolinghe and Wudang Groups of low-grade metamorphic volcanosedimentary rock assemblages [45]. In addition, Neoproterozoic tillite and platform carbonate, Cambrian-Ordovician limestone, and Silurian shale unconformably overlie the basements.

Analytical Methods
Apatites were extracted separately from carbonatite, syenite, and hornblendite, following standard density and magnetic separation procedures. Apatite grains were then handpicked, mounted in epoxy resin, and polished to expose the crystals' interiors. Polished mounts were then documented with photomicrographs before analyses to characterize apatite grains' crystal morphology ( Figure 2g) and select targets for in situ elemental and Sr-Nd isotope analyses.

Major and Trace Elements
Major element compositions were measured using a JEOL-JAX8100 electron microprobe (EPMA) at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS), Beijing, China. The operating conditions were 15 kV accelerating voltage, 20 nA beam current, and 5 µm beam diameter. The standards for apatite analyses were apatite for Ca and P, albite for Na, diopside for Si, barite for S and Ba, celestine for Sr, monazite for La, Ce, Pr, and Nd, Mn oxide (MnO) for Mn, tugtupite for Cl, and synthetic fluorite for F. Apatite formulas were normalized to eight cations, assuming stoichiometry. Detection limits for analyzed elements ranged from ca. 100 ppm to 200 ppm.
Trace elements were determined on polished grain mounts of apatite by laser ablation inductively coupled plasma mass spectrometry (LA-MC-ICP-MS) equipped with a 193 nm excimer ArF LA system at the IGGCAS, with a 1 mm Ni sample cone and 0.4 mm Ni skimmer cone. The ablation cell was connected to an Agilent 77500a ICP-MS. Detailed instrument operation conditions, analytical procedures, and data reduction are described by Wu et al. [49]. To analyze these apatite grains, we used 40-60 µm laser craters pulsed at 6 Hz and an energy fluence of 6 J/cm 2 . Dwell times were 10 ms for U and Th and 6 ms for isotopes of other elements; analyses involved~30 s background acquisition and 50 s sample data acquisition, and helium was used as the carrier gas at 750 mL/ min. NIST 610 was used as the primary external calibration standard, and 43 Ca was used for internal calibration. Data were processed using the GLITTER program (GEMOC, Macquarie University). Relative analytical uncertainties were within 20% for Rb, Zr, Hf, and Ta, and 10% for other trace elements.

In Situ Sr and Nd Isotopic Composition
In situ Sr and Nd isotopic analyses of apatite were undertaken using a Thermo Scientific Neptune MC-ICP-MS and 193 nm excimer ArF LA system at the IGGCAS. The instrumental operation, analytical procedures, and data reduction are discussed in Yang et al. [50][51][52]. For Sr analysis, a laser repetition rate of 8 Hz with an energy fluence of 15 J/cm 2 was applied; the spot size was 80-120 µm. The influence of Ca on Sr isotope analysis was insignificant. The 87 Sr/ 86 Sr ratios obtained for in-house standard apatites AP1 and Slyudyanka were 0.711300 ± 0.000092 (2SD; n = 64) and 0.707752 ± 0.000010 (2SD; n = 44), respectively, consistent with their reference values [52].
With respect to the Nd isotope, a laser repetition rate of 6 Hz with an energy fluence of 15 J/cm 2 was applied with spot sizes of 90 µm. The influence of 142 Ce on 142 Nd was insignificant, so the main isobaric interference for Nd was that of 144 Sm on 144 Nd, which was corrected with a procedure similar to that of McFarlane et al. [53]. The 147 Sm/ 144 Nd and 143 Nd/ 144 Nd ratios obtained for in-house standard apatites MAD and Otter Lake were 0.082703 ± 0.000091 (2SD; n = 64) and 0.511348 ± 0.0000054 (2SD; n = 95), and 0.08472 ± 0.00039 and 0.511955 ± 0.0000064, respectively, which are identical to values obtained by solution methods and long-term determinations by LA-MC-ICP-MS [52,54].

Sr and Nd Isotopic Composition
The Sr-Nd isotopic compositions of apatite are listed in Table 3 and Supplementary Materials S3. The initial isotopic ratios of the apatites in the SXD carbonatite are calculated back to 220Ma based on our unpublished dating results of biotite, yielding initial 87 Sr/ 86 Sr ratios of 0.70202-0.70332 and ε Nd (t) values of 0.6-3.4 (Figure 4a). For SXD hornblendite and syenite, the initial Sr and Nd isotopic ratios were recalculated based on the zircon U-Pb age (435 Ma, our unpublished data). The initial 87 [55] and Sun and McDonough [56], respectively. The REE data of calcite are from Xu et al. [37] and Chen et al. [38].

Sr and Nd Isotopic Composition
The Sr-Nd isotopic compositions of apatite are listed in Table 3 and Supplementary Materials S3. The initial isotopic ratios of the apatites in the SXD carbonatite are calculated back to 220Ma based on our unpublished dating results of biotite, yielding initial 87 Sr/ 86 Sr ratios of 0.70202-0.70332 and ɛNd(t) values of 0.6-3.4 (Figure 4a). For SXD hornblendite and syenite, the initial Sr and Nd isotopic ratios were recalculated based on the zircon U-Pb age (435 Ma, our unpublished data). The initial 87 Sr/ 86 Sr and ɛNd(t) are 0.70301-0.70333 and 1.5-3.8, and 0.70337-0.70389 and 2.5-6.2 for hornblendite and syenite apatite, respectively (Figure 4b).   [46]. The East African carbonatite line (EACL) and Oldoinyo Lengai are referred to in [57][58][59]. The normal lithospheric mantle is constrained by the Early Paleozoic mafic dykes in South Qinling [60,61]. The Yangtze upper/middle crust data are from Chen and Jahn [62], Gao et al. [63], and Ma et al. [64]. The data of the early Paleozoic mafic-ultramafic rocks in the South Qinling Belt are from [65].

Origin of the Apatite: Magmatic or Metasomatic?
Apatite, which occurs throughout early-stage magmatic to late-stage hydrothermal environments, has distinct morphology and geochemical signatures [66]. Apatite of magmatic origin is generally euhedral to subhedral with or without zoning textures [18,21].  [46]. The East African carbonatite line (EACL) and Oldoinyo Lengai are referred to in [57][58][59]. The normal lithospheric mantle is constrained by the Early Paleozoic mafic dykes in South Qinling [60,61]. The Yangtze upper/middle crust data are from Chen and Jahn [62], Gao et al. [63], and Ma et al. [64]. The data of the early Paleozoic mafic-ultramafic rocks in the South Qinling Belt are from [65]. Apatite, which occurs throughout early-stage magmatic to late-stage hydrothermal environments, has distinct morphology and geochemical signatures [66]. Apatite of magmatic origin is generally euhedral to subhedral with or without zoning textures [18,21]. They typically contain F as the dominant volatile species over Cl and H 2 O [67]. Apatites in various kinds of host igneous rocks are highly diverse in terms of trace element composition [1,19,27]. For example, apatites in granite are characterized by a strong negative Eu anomaly, resulting mainly from plagioclase crystallization [68,69]. As the common REE-enriched mineral in carbonatite, alkaline, mafic, and ultramafic rocks, apatite is enriched in LREEs with high (La/Yb) N and Sr/Y ratios, corresponding to the right-sloping patterns in the chondrite-normalized REE diagram [18,21,27]. Metasomatic apatites usually show turbid cores and overgrowth rims with abundant inclusions of xenotime and monazite. They have relatively lower REE concentrations. The preferential transport of the LREEs away from the original apatite leads to passive enrichment of HREEs in the metasomatic apatites, leaving low (La/Yb) N ratios and flat or even left-sloping patterns in the chondrite-normalized REE diagram [3,12,48]. Hydrothermal apatites occur as fine anhedral grains and show association with baryte, fluorite, and REE minerals (e.g., synchysite and monazite). Their REE patterns are highly variable [18,[70][71][72][73][74] due to the control of REE transport and fractionation by REE complexes of CO 2 and saline (chloride or fluoride) in hydrothermal fluids [75][76][77][78][79].
As shown in Figure 2a,b, the apatite (<50 µm) from the SXD carbonatite occurs as fine anhedral inclusions in calcite and pyrite, suggesting that apatite can crystallize at early magmatic stages. The apatite grains studied are euhedral to subhedral (>80 µm) with a well-formed crystal shape (Figure 2g), irrespective of lithology. They have clear cores and lack monazite inclusions. They are all typical fluorapatite as evidenced by high F and undetected Cl contents. They are enriched in REEs and have high Sr/Y ratios (Figure 5a), indicative of magmatic origin and little or no metasomatic influence [27]. They have high (La/Sm) N and (La/Yb) N ratios, falling into the field of magmatic apatite and showing a sharp contrast with the altered apatites from the SXD complex ( Figure 5b) [48,70]. Moreover, altered apatites are usually characterized by inclusions of xenotime and monazite [48], which is not the case for the apatites we studied. In addition, they also show similar compositions with the magmatic apatite of corresponding rock types (Figure 5c-e). The Sr and Nd isotopic compositions of the apatites are homogenous. They overlap those of their host rocks (Figure 4), indicating that the apatite had been fractionally crystallized from the host magma that had evolved in a closed system. Trace element abundances of the magmatic apatite are mainly controlled by partition coefficients, geochemistry of the host magma, and crystallization phases simultaneous to or before the apatite. Thus, their geochemical signatures are used to decipher the evolution of the parental magmas and the nature of their mantle sources.

Carbonatite
Apatite is the primary REE-bearing accessory mineral in carbonatite. Previous studies have demonstrated that REE substitutes on the Ca site accompanied by Na substitution on the Ca site or Si substitution on the P site can compensate for charge excess [15,17,19,83]. The apatite from the SXD carbonatite shows low Ba, Si, S, and Mn contents, and there is no apparent correlation between them. By contrast, Na, Sr, and LREEs correlate well with the P and Ca (Figure 6a). In general, these elements are preferred to substitute for Ca [15][16][17]. Thus, the substitution formula is likely to be: and Nd isotopic compositions of the apatites are homogenous. They overlap those of their host rocks (Figure 4), indicating that the apatite had been fractionally crystallized from the host magma that had evolved in a closed system. Trace element abundances of the magmatic apatite are mainly controlled by partition coefficients, geochemistry of the host magma, and crystallization phases simultaneous to or before the apatite. Thus, their geochemical signatures are used to decipher the evolution of the parental magmas and the nature of their mantle sources.  [18,21,42,48,70,[80][81][82] and references therein. The primary and altered apatite in the SXD complex are referred to in [48,70]. Data for apatite in carbonatite, alkaline rock, and mafic/ultramafic rocks are from [21] and references therein.

Carbonatite
Apatite is the primary REE-bearing accessory mineral in carbonatite. Previous studies have demonstrated that REE substitutes on the Ca site accompanied by Na substitution on the Ca site or Si substitution on the P site can compensate for charge excess [15,17,19,83]. The apatite from the SXD carbonatite shows low Ba, Si, S, and Mn contents, and there is no apparent correlation between them. By contrast, Na, Sr, and LREEs correlate well with the P and Ca (Figure 6a). In general, these elements are preferred to substitute for Ca [15][16][17]. Thus, the substitution formula is likely to be: (1)  [18,21,42,48,70,[80][81][82] and references therein. The primary and altered apatite in the SXD complex are referred to in [48,70]. Data for apatite in carbonatite, alkaline rock, and mafic/ultramafic rocks are from [21] and references therein. The apatites in SXD carbonatite show high total REE contents (up to 15000 ppm). Experimental results reveal that REEs are moderately incompatible in apatite in the pure carbonatitic melt without silica contents [84], and their compatible behaviors are independent of temperature, pressure, or volatiles [84,85]. However, increasing evidence demonstrates that the addition of silica elevates the polymerization of the carbonatite melt, which significantly enhances the partition coefficients of REEs between apatite and melt by means of facilitating the entry of REEs into the Ca site of apatite [13,18,42,86,87]. It is clear that the SXD carbonatite is not pure, as evidenced by the whole-rock SiO2 content of 2.7% on average and the occurrence of modal silicate minerals, such as biotite. So, the high REE contents in SXD carbonatite apatite can be reasonably accounted for by the elevated silica in the SXD carbonatite.
As a result of fractionation of REE-enriched apatite, the residual melt should be depleted of REEs and have a lower enrichment of La with respect to Nd (i.e., LREE/HREE) [18], which is consistent with the relatively flat chondrite-normalized REE patterns of the calcite (Figure 3a) [37,38]. In addition, (La/Yb)N and Th are positively correlated with LREEs (Figure 7a,c), which is probably due to the crystallization of pyrochlore or allanite in magma, as both are enriched in LREEs. The difference between allanite and pyrochlore is that the former is strongly enriched in Th over U. Therefore, the crystallization of allanite would lead to an extreme depletion of Th with a slight decrease of U in the residual The apatites in SXD carbonatite show high total REE contents (up to 15000 ppm). Experimental results reveal that REEs are moderately incompatible in apatite in the pure carbonatitic melt without silica contents [84], and their compatible behaviors are independent of temperature, pressure, or volatiles [84,85]. However, increasing evidence demonstrates that the addition of silica elevates the polymerization of the carbonatite melt, which significantly enhances the partition coefficients of REEs between apatite and melt by means of facilitating the entry of REEs into the Ca site of apatite [13,18,42,86,87]. It is clear that the SXD carbonatite is not pure, as evidenced by the whole-rock SiO 2 content of 2.7% on average and the occurrence of modal silicate minerals, such as biotite. So, the high REE contents in SXD carbonatite apatite can be reasonably accounted for by the elevated silica in the SXD carbonatite.
As a result of fractionation of REE-enriched apatite, the residual melt should be depleted of REEs and have a lower enrichment of La with respect to Nd (i.e., LREE/HREE) [18], which is consistent with the relatively flat chondrite-normalized REE patterns of the calcite ( Figure 3a) [37,38]. In addition, (La/Yb) N and Th are positively correlated with LREEs (Figure 7a,c), which is probably due to the crystallization of pyrochlore or allanite in magma, as both are enriched in LREEs. The difference between allanite and pyrochlore is that the former is strongly enriched in Th over U. Therefore, the crystallization of allanite would lead to an extreme depletion of Th with a slight decrease of U in the residual melt [88]. In the case of SXD carbonatite, apatite displays extremely low U (~0.1 ppm) but variable Th (0.3-37 ppm) [85], which precludes allanite as a significant crystallization phase. Pyrochlore is the dominant Nb-bearing accessory mineral in the SXD carbonatite. Its euhedral hexagonal crystal shape suggests that it crystallized during early-stage magmatism. As pyrochlore usually has high U (>1300 ppm) and Th (>200 ppm) contents [38], crystallization of pyrochlore would decrease Nb, Th, U, and LREE contents in the residual melt, consistent with the strong negative Nb anomaly of apatite and calcite (Figures 3b and 7c,d) [37,38].

Syenite
The apatites from the SXD syenite have high Na2O, SrO, and LREEs (Figure 6a), which are preferred to substitute for Ca [15][16][17], and the substitution formula may be: The relatively high Na2O contents of apatites suggest that they should have crystallized from a Na-rich magma [87], which is consistent with the high Na2O (8.8-9.8 wt%) of their host syenite. Experimental studies reveal that the REEs are compatible in apatite crystallizing from silicate melts, and middle REEs (MREEs) usually have higher partition coefficients than LREEs and HREEs ( ap/liq DMREE> ap/liq DLREE or ap/liq DHREE) [83,87,89]. This is consistent with the concave-downward REE patterns of the apatite from the syenite (Figure 3c). The extremely low Zr, Hf, and U contents compared with the host syenite are indicative of the coeval crystallization of zircon. Generally, the partitioning of Sr between apatite and silicate melt is relatively constant and insensitive to changes in temperature, pressure, volatiles, or melt composition. Thus, the positive correlation between LREE and Sr contents (Figure 7b) suggests that the partition coefficients of REEs are relatively constant as well. Therefore, the positive correlation between LREEs and (La/Yb)N (Figure 7a) is likely attributed to the crystallization of LREE-rich minerals, such as monazite and allanite. As allanite is enriched in Th and LREEs, its crystallization would decrease LREEs and Th in the melt. Thus, the decrease of LREEs with decreasing Th (Figure 7c) shown by the apatite in syenite points to a potential coeval crystallization of allanite rather than

Syenite
The apatites from the SXD syenite have high Na 2 O, SrO, and LREEs (Figure 6a), which are preferred to substitute for Ca [15][16][17], and the substitution formula may be: The relatively high Na 2 O contents of apatites suggest that they should have crystallized from a Na-rich magma [87], which is consistent with the high Na 2 O (8.8-9.8 wt%) of their host syenite. Experimental studies reveal that the REEs are compatible in apatite crystallizing from silicate melts, and middle REEs (MREEs) usually have higher partition coefficients than LREEs and HREEs ( ap/liq D MREE > ap/liq D LREE or ap/liq D HREE ) [83,87,89]. This is consistent with the concave-downward REE patterns of the apatite from the syenite (Figure 3c). The extremely low Zr, Hf, and U contents compared with the host syenite are indicative of the coeval crystallization of zircon. Generally, the partitioning of Sr between apatite and silicate melt is relatively constant and insensitive to changes in temperature, pressure, volatiles, or melt composition. Thus, the positive correlation between LREE and Sr contents (Figure 7b) suggests that the partition coefficients of REEs are relatively constant as well. Therefore, the positive correlation between LREEs and (La/Yb) N (Figure 7a) is likely attributed to the crystallization of LREE-rich minerals, such as monazite and allanite. As allanite is enriched in Th and LREEs, its crystallization would decrease LREEs and Th in the melt. Thus, the decrease of LREEs with decreasing Th (Figure 7c) shown by the apatite in syenite points to a potential coeval crystallization of allanite rather than monazite, which is also consistent with the absence of monazite in SXD syenite.

Hornblendite
The apatites from the hornblendite have relatively high Na, Si, and S, but low Sr contents. They show an insignificant correlation between P + Ca and REE + Na, indicating a different substitution mechanism from those in apatite for both carbonatite and syenite (Figure 6a). There is a perfect negative correlation between P + Ca and REE + Si + S + Na ( Figure 6b) and a positive correlation between S + REE and Si and Na. Thus, the substitution for the apatite from the SXD hornblendite is likely to be: It has been well demonstrated that apatite from most mafic rocks shows smooth or concave-downward REE patterns due to the high partition coefficients of MREEs [27,90,91]. Apatites from the SXD hornblendite exhibit weak concave-upward REE patterns with slight negative Eu anomalies (δEu = 0.73-0.95, Figure 8). They usually occur interstitially, implying crystallization later than the cumulus amphibole. Since the amphibole is invariably characterized by higher MREE partition coefficients relative to the LREEs and HREEs, their cumulate would leave an MREE-decreased residual melt, from which the apatite crystallized subsequently. That may be the main cause of the weak concave-upward chondrite-normalized REE patterns in the SXD hornblendite apatite (Figure 3e). The SXD hornblendite whole rock has a negligible Eu anomaly (0.96) compared to the apatite. As there is no plagioclase in the SXD hornblendite, the negative Eu anomaly of apatite is not caused by plagioclase fractionation. Ce and Eu generally have two ionic valences each: Ce 3+ , Ce 4 , Eu 2+ , and Eu 3+ . Ce 3+ and Eu 3+ preferentially substitute for Ca 2+ in apatite due to their similar ionic radii [15,16,19,92]. Therefore, the weak negative correlation between δEu and δCe most likely indicates more Ce 3+ and Eu 2+ relative to Ce 4+ and Eu 3+ in the host magma, suggesting that magma evolution of SXD hornblendite may have proceeded under relatively reduced conditions ( Figure 8) [93]. Calcic amphibole (kaersutite and magnesiohastingsite) may incorporate more Eu 2+ , as their Ca 2+ has a similar ionic radius with that of Eu 2+ [94][95][96][97].  Figure 6a). There is a perfect negative correlation between P + Ca and REE + Si + S+Na ( Figure 6b) and a positive correlation between S + REE and Si and Na. Thus, the substitution for the apatite from the SXD hornblendite is likely to be: It has been well demonstrated that apatite from most mafic rocks shows smooth or concave-downward REE patterns due to the high partition coefficients of MREEs [27,90,91]. Apatites from the SXD hornblendite exhibit weak concave-upward REE patterns with slight negative Eu anomalies (δEu = 0.73-0.95, Figure 8). They usually occur interstitially, implying crystallization later than the cumulus amphibole. Since the amphibole is invariably characterized by higher MREE partition coefficients relative to the LREEs and HREEs, their cumulate would leave an MREE-decreased residual melt, from which the apatite crystallized subsequently. That may be the main cause of the weak concave-upward chondrite-normalized REE patterns in the SXD hornblendite apatite ( Figure  3e). The SXD hornblendite whole rock has a negligible Eu anomaly (0.96) compared to the apatite. As there is no plagioclase in the SXD hornblendite, the negative Eu anomaly of apatite is not caused by plagioclase fractionation. Ce and Eu generally have two ionic valences each: Ce 3+ , Ce 4 , Eu 2+ , and Eu 3+ . Ce 3+ and Eu 3+ preferentially substitute for Ca 2+ in apatite due to their similar ionic radii [15,16,19,92]. Therefore, the weak negative correlation between δEu and δCe most likely indicates more Ce 3+ and Eu 2+ relative to Ce 4+ and Eu 3+ in the host magma, suggesting that magma evolution of SXD hornblendite may have proceeded under relatively reduced conditions ( Figure 8) [93]. Calcic amphibole (kaersutite and magnesiohastingsite) may incorporate more Eu 2+ , as their Ca 2+ has a similar ionic radius with that of Eu 2+ [94][95][96][97].

Implication for the Mantle Source
The magmatic apatites from the SXD complex show Sr-Nd isotopic compositions overlapping with their host rocks, arguing for magma evolution in a closed system. Thus, the isotopic ratios of apatites are representative of their mantle sources.
During the early Paleozoic, the SQB was in a rift setting, as evidenced by widespread

Implication for the Mantle Source
The magmatic apatites from the SXD complex show Sr-Nd isotopic compositions overlapping with their host rocks, arguing for magma evolution in a closed system. Thus, the isotopic ratios of apatites are representative of their mantle sources.
During the early Paleozoic, the SQB was in a rift setting, as evidenced by widespread collapsed basins, faults, and rift-related igneous rocks [43,45]. Numerous mafic dykes and sills of similar age to the SXD hornblendite and syenite are widely distributed in the North Daba Mountains, Wudang terrane, and Suizao area [65,98]. They are all of the alkaline affinity and are characterized by OIB-like trace element signatures [61,98,99]. The Sr-Nd-Pb isotopic compositions reveal that their mantle source follows a mixing trend between DMM/HIMU and EM (EMI, EMII, or both) components [99]. Accumulating geochemical evidence and studies of pyroxenite-hornblendite xenoliths further point to a metasomatic lithospheric mantle [46,[100][101][102]. Considering the geochemical similarity, especially the similar Sr and Nd isotopic compositions between the SXD hornblendite and syenite and the early Paleozoic mafic-ultramafic rocks in the SQB (Figure 4b), it is reasonable to postulate that the SXD hornblendite and syenite may share a common metasomatic lithospheric mantle source with these early Paleozoic mafic-ultramafic rocks.
Carbonatite is considered the product of low-degree partial melting of the mantle and is commonly associated with extensional settings [103]. Thus, the formation of SXD carbonatites most likely suggests a shift from a compressional to a post-collision extensional regime at ca. 220Ma. There are two types of Triassic mafic dykes occurring in the Wudang terrane and they are considered to be derived from the asthenosphere (Mesozoic dyke II in Figure 4b) and a mixed mantle source between the asthenosphere and subducted crustal material (Mesozoic dyke I in Figure 4b), respectively [46]. It is clear that the SXD carbonatite is different from these two types of dykes in terms of Sr and Nd isotopic compositions, implying a highly heterogeneous mantle beneath the SQB during the Triassic. The Sr-Nd isotopic compositions of the apatite in SXD carbonatite fall into the field defined by Oldoinyo Lengai carbonatite and on the East African carbonatite line (EACL), which are considered to be connected to the HIMU and EM I mantle components [57][58][59]. Diverse components have been proposed to account for the HIMU signature, such as the recycled basaltic oceanic crust [104], metasomatic lithospheric mantle [105,106] or Archean to early Proterozoic subduction-related carbonatite-metasomatized sub-continental lithospheric mantle [107][108][109]. In the SXD case, however, the carbonatite has a lighter Mg isotopic composition than the normal mantle (our unpublished data), unequivocally demonstrating the involvement of recycled carbonate in its mantle source. Considering the diffusion coefficient of the Mg isotope in inter-mineral systems [110,111], the recycled crustal materials of Archean or Proterozoic would be re-equilibrated with the surrounding mantle. Thus, we propose that sedimentary carbonates, subducted following the final closure of the Mianlue ocean and subsequent amalgamation of the SQB and SCB in the Triassic, would be the proper candidates to account for the light-Mg isotope signature. Therefore, we propose that the SXD carbonatite may be derived from a mixed mantle source of HIMU components and subducted sedimentary carbonates.

1.
Apatites from the SXD complex are of magmatic origin and have not experienced subsequent metasomatism or other alteration. Their Sr-Nd isotopic compositions that overlap with host rocks indicate magma evolution in a relatively closed system with no significant crustal contamination.

3.
Pyrochlore crystallization plays an important role in controlling the REE signature of apatite from the carbonatite, while the fractionation of zircon, allanite, and amphi-