Post-Magmatic Fluids Dominate the Mineralization of Dolomite Carbonatitic Dykes Next to the Giant Bayan Obo REE Deposit, Northern China

: The Bayan Obo rare earth element (REE) deposit in Inner Mongolia, northern China, is the largest REE deposit in the world, whose mineralization process remains controversial. There are dozens of carbonatite dykes that are tightly related to the deposit. Here we report the petrological and mineralogical characteristics of a typical dolomite carbonatite dyke near the deposit. The dolomite within the dyke experienced intense post-emplacement ﬂuids metasomatism as evidenced by the widespread hydrothermal REE-bearing minerals occurring along the carbonate mineral grains. REE contents of bulk rocks and constituent dolomite minerals ( > 90 vol.%) are 1407–4184 ppm and 63–152 ppm, respectively, indicating that dolomite is not the dominant mineral controlling the REE budgets of the dyke. There are three types of apatite in the dyke: Type 1 apatite is the primary apatite and contains REE 2 O 3 at 2.35–4.20 wt.% and SrO at 1.75–2.19 wt.%; Type 2 and Type 3 apatites are the products of replacement of primary apatite. The REE 2 O 3 (6.10–8.21 wt.%) and SrO (2.83–3.63 wt.%) contents of Type 2 apatite are signiﬁcantly elevated for overprinting of REE and Sr-rich ﬂuids derived from the carbonatite. Conversely, Type 3 apatite has decreased REE 2 O 3 (1.17–2.35 wt.%) and SrO (1.51–1.99 wt.%) contents, resulting from inﬁltration of ﬂuids with low REE and Na concentrations. Our results on the dyke suggest that post-magmatic ﬂuids expelled from the carbonatitic melts dominated the REE mineralization of the Bayan Obo deposit, and a signiﬁcant ﬂuid disturbance occurred but probably provided no extra REEs to the deposit.


Analytical Methods
The detailed microscopic observations of thin sections and mineral separates using transmitted light, scanning electron microscope (SEM, Nova Nano SEM450, FEI, Hillsboro, OR, USA) and cathodoluminescence (CL) were carried out prior to further geochemical analysis. Major and trace element analysis of five dolomite rock samples were conducted at the Geoanalytical Center of Nuclear Industry, Beijing. The major elements were determined by X-ray fluorescence spectrometry (Axios mAX, PANalytical B.V., Almelo, Netherlands) with analytical precision better than 2%. Trace elements were determined using an inductively coupled plasma mass spectrometer (ICP-MS, NexION300D, PerkinElmer, Waltham, MA, USA) with analytical precision better than 5%. Before analysis, samples were dissolved in a mixture of HF + HNO 3 within a closed Teflon vessel on a hotplate at about 160 • C for 48 h and evaporated to dryness. After that, HClO 4 was added to the vessel and heated again to about 160 • C until the acid was completely evaporated. After cooling to room temperature, the dry sample was then dissolved in 1% HNO3 to attain the solution for trace element determinations.
Major element compositions of apatite were determined by electron probe microanalyser (EPMA) using the Jeol JXA-8100 (Jeol Ltd., Tokyo, Japan) at Geoanalytical Center of Nuclear Industry, Beijing. An accelerating voltage of 15 kV, beam size of 10 µm, and beam current of 20 nA were adopted for the analysis. Natural and synthetic materials were used as standards that were previously tested for homogeneity before utilization for quantitative analysis. The standards used were BaF 2 for F, fluorapatite for Ca and P, and monazite for REEs. Matrix corrections were conducted by using the ZAF correction program supplied by the manufacturer.
Trace elements of dolomite were obtained using a New Wave UP 213 laser ablation system coupled to a Thermo-Finnigan Element II ICP-MS (Thermo Fisher Scientific, Waltham, MA, USA) at the National Research Center for Geoanalysis, Beijing. The laser was operated at a spot size of 40 µm in diameter and 12 J/cm 2 energy beam at 10 Hz. He was used as sweeping gas. Analyses of the standards (NIST610, GSE-1G, MACS-3) were conducted 3 times before and after each run of 10 sample analyses. The preferred values of element concentrations for the United States Geological Survey (USGS) reference glasses were taken from the GeoReM database (http://georem.mpch-mainz.gwdg.de/). The raw data were processed using Microsoft Excel 2007. The MACS-3 and NIST610 were used for external calibration and Ca ion signal intensities were employed as the internal standard to monitor instrumental drift [37]. The content (wt.%) of CaO was obtained using EPMA. The detection limit for most elements was considerably below 0.1 ppm.

Petrography
The rock samples consist of predominantly (>90% Vol) subhedral medium-grained dolomite with sizes of −500 µm, and accessory minerals (<5% Vol) including calcite, magnetite, and apatite. Primary monazite or bastnaesite intergrown with dolomite is rare. The magnetite occurs as irregular grains disseminated along dolomite grain boundaries (Figure 3b). Besides magnetite, dark anhedral minerals also occur as aggregates along the dolomite grain (Figure 3b,c) with fibrous or micro-granular textures, which are clearly the products of hydrothermal fluids after the crystallization of the dolomite. Energy dispersive spectrometry (EDS) analysis indicates that they are parasite and barite, as well as some Nb-, Fe-, Sr-and Ba-bearing minerals that cannot be identified (Figure 3d-f).
The apatite grains are euhedral with grain sizes of 100-200 µm. They commonly have smooth and flat contacts with surrounding dolomites (Figure 4a,b). BSE and CL images revealed that there are three types of apatite in the dyke rocks. Apart from apatite grains (Type 1) with relatively smooth surfaces and margins (Figure 4b

Bulk Rock Geochemistry
Major and trace element compositions for five dolomite rock samples are listed in Table 1

Bulk Rock Geochemistry
Major and trace element compositions for five dolomite rock samples are listed in Table 1 Figure 5). P 2 O 5 contents are relatively low in these samples (0.08-0.37 wt.%). In addition, the samples have high Sr (13,800-18,400 ppm) and Ba (1198-5833 ppm) contents. Their primitive mantle-normalized trace element patterns show positive anomalies of Ba, Sr, and Nb and negative anomalies of Zr, Hf, Ta, and Pb ( Figure 6a). The REE contents of these samples range from 1407 to 4184 ppm (2243 ppm on average) with (La/Yb) N ratios of 243-1356. Their chondrite-normalized REE patterns display a strong enrichment in light rare earth element (LREE) and weak Eu negative anomalies ( Figure 6b).   for rock samples from the Dol-dyke. The gray lines are data of ore-hosting dolomites from the Bayan Obo deposit [26]. Data for normalisation according to [39].

Mineral Chemistry of Dolomite and Apatite
Trace element compositions for dolomite grains are listed in Table 2. As the major constituent mineral of the dyke, dolomite contains total REE contents of 62.8-152.2 ppm (100.3 ppm on average), which is much less than those of the bulk rock and the coarse-grained dolomite of the Bayan Obo orehosting rock reported by Chen et al. [26]. Their chondrite-normalized REE patterns are relatively flat ( Figure 7) with (La/Yb)N ratios range from 49 to 100. In addition, they have Sr and Ba contents of 5120-6036 ppm and 72-98 ppm, respectively.  for rock samples from the Dol-dyke. The gray lines are data of ore-hosting dolomites from the Bayan Obo deposit [26]. Data for normalisation according to [39].

Mineral Chemistry of Dolomite and Apatite
Trace element compositions for dolomite grains are listed in Table 2. As the major constituent mineral of the dyke, dolomite contains total REE contents of 62.8-152.2 ppm (100.3 ppm on average), which is much less than those of the bulk rock and the coarse-grained dolomite of the Bayan Obo orehosting rock reported by Chen et al. [26]. Their chondrite-normalized REE patterns are relatively flat ( Figure 7) with (La/Yb)N ratios range from 49 to 100. In addition, they have Sr and Ba contents of 5120-6036 ppm and 72-98 ppm, respectively. for rock samples from the Dol-dyke. The gray lines are data of ore-hosting dolomites from the Bayan Obo deposit [26]. Data for normalisation according to [39].

Mineral Chemistry of Dolomite and Apatite
Trace element compositions for dolomite grains are listed in Table 2. As the major constituent mineral of the dyke, dolomite contains total REE contents of 62.8-152.2 ppm (100.3 ppm on average), which is much less than those of the bulk rock and the coarse-grained dolomite of the Bayan Obo ore-hosting rock reported by Chen et al. [26]. Their chondrite-normalized REE patterns are relatively flat ( Figure 7) with (La/Yb) N ratios range from 49 to 100. In addition, they have Sr and Ba contents of 5120-6036 ppm and 72-98 ppm, respectively.   [26]. Data for normalization according to [39].
The major element compositions for apatite are listed in Table 3

Formation of the Three Types of Apatite
As mentioned earlier, the three types of apatite show similar textures, grain sizes, and P 2 O 5 and F contents, indicating that they were originally a batch of crystallized products from carbonatitic melts. And for all types of apatite in the dyke, CaO contents display negative correlations with LREE 2 O 3 + Na 2 O and SrO contents, and there is a significant positive correlation between LREE 2 O 3 and Na 2 O contents (Figure 8). These indicate that for all apatite grains in the dyke, the REE and Sr go in and out of the apatite mainly via the coupled substitution Na + + REE 3+ = 2Ca 2+ and Sr 2+ = Ca 2+ . Among the mineral constituents of the rock samples, apatite seems to be the main gauge of fluid interaction. REE and Sr in apatite can be easily affected by hydrothermal alteration processes [40][41][42][43]. Therefore, tracing these element variations in apatite is a good way to understand the crystallization and overprinted subsequent processes. In this study, the three types of apatite exhibit distinctly different REE and Sr contents. As shown in Figure 9, Type 1 apatite has LREE 2 O 3 contents comparable to primary apatite in other Dol-dykes [44] from the Bayan Obo ore district. BSE and CL images of Type 1 display no reaction textures, indicating that it was not affected or was little affected by post-crystallization alteration. Therefore, it can represent the primary apatite crystallized from carbonatitic melts.     According to reaction textures and compositional changes, Type 2 and Type 3 are the products of replacement of the primary apatite (Type 1 apatite). Compared with Type 1 apatite, Type 2 apatite has a sharp increase in contents of LREE 2 O 3 , Na 2 O, and SrO. In fact, Type 2 apatite contains the highest REE contents found so far in the Bayan Obo ore district. This, in conjunction with the monazite replacements at the Type 2 apatite grain boundaries or internal fractures, indicates an interaction between primary apatite and LREE-, Na-, and Sr-rich fluids. The process can be represented by Reaction (1): Type 1 apatite + (LREE 3+ , Na + , Sr 2+ ) in fluids = Type 2 apatite + monazite + Ca 2+ in fluids. The X-ray element scanning images ( Figure 10) show that the Type 2 apatite is relatively homogeneous in Ca, P, and Sr compositions, which means a complete re-equilibration between Type 2 apatite and the extensive fluids. The formation of metasomatic monazites along the apatite grain boundaries or interior fractures indicates that phosphate probably had an important role in the precipitation of REE in fluids.
Type 3 apatite shows dark CL images as compared with the bright CL images of Type 1 apatite (Figure 4g,h). This luminescent intensity difference can be explained by the REE variation [45]. Under the same conditions, apatite that has been metasomatized by fluids and consequently lost LREEs appears dark in CL images [46]. In this study, Type 3 apatite contains lower total LREE 2 O 3 , Na 2 O, and SrO contents than Type 1 apatite. The veinlets in Type 3 apatite are the main argument that this is an overprinted apatite and not a primary apatite, and they are also direct evidence for the mobility of hydrothermal fluids. Therefore, it is obvious that the leaching of LREE from primary apatite occurred. The migration of LREE out of apatite can be expressed by Reaction (2): Type 1 apatite + Ca 2+ in fluids = Type 3 apatite + LREE 3+ , Sr 2+ and Na + in fluids.
Reaction (2) can be regarded as the reverse of reaction (1), so the nature of the fluids in Reaction (2) should be completely different from that of fluids in Reaction (1). Besides having low LREE and Sr concentrations to impel the leaching process, the fluids in Reaction (2) were also supposed to be poor in Na or Si contents, because these elements can freely enter the apatite lattices to maintain an electrovalent balance, thereby inhibiting the removal of REE from apatite [47,48].
In conclusion, primary apatite in the dyke was affected by fluids at two separate stages: a first metasomatic stage represented by the formation of Type 2 apatite, driven by post-magmatic fluids with fairy high LREE, Na, and Sr contents derived from evolved carbonatitic magmas, and a second stage of metasomatic modification caused LREEs leaching from the primary apatite and formed Type 3 apatite. In conclusion, primary apatite in the dyke was affected by fluids at two separate stages: a first metasomatic stage represented by the formation of Type 2 apatite, driven by post-magmatic fluids with fairy high LREE, Na, and Sr contents derived from evolved carbonatitic magmas, and a second stage of metasomatic modification caused LREEs leaching from the primary apatite and formed Type 3 apatite.

The REE Mineralization of the Dol-Dyke
As noted earlier, dolomite mineral comprises more than 90% of the sample's volume but contains very low REE contents (62.8-152.2 ppm). Hence, it contributes little to the whole-rock REE budget (1407-4184 ppm) of the dyke. Due to the lack of primary monazite or bastnaesite, the main primary REE-bearing mineral in the dyke is apatite. However, the low P2O5 content (0.23 wt.% on average) of the rock samples indicates that apatite only comprises a very small proportion. Therefore, the primary of apatite role in controlling the whole-rock REE budget can be excluded. As a result, the REE contents of the dyke come from the hydrothermal REE-bearing minerals. That is, hydrothermal fluids dominate the REE mineralization of the dyke. The fluids shall be the same as the post-magmatic fluids responsible for the formation of Type 2 apatite mentioned above.
Bulk rock REE contents of other carbonatite dykes in the ore district have also been reported in previous studies, and the data vary greatly between different types of dykes, different dykes of the same type, or different zones of the same dyke. For example, REE contents of rock samples from

The REE Mineralization of the Dol-Dyke
As noted earlier, dolomite mineral comprises more than 90% of the sample's volume but contains very low REE contents (62.8-152.2 ppm). Hence, it contributes little to the whole-rock REE budget (1407-4184 ppm) of the dyke. Due to the lack of primary monazite or bastnaesite, the main primary REE-bearing mineral in the dyke is apatite. However, the low P 2 O 5 content (0.23 wt.% on average) of the rock samples indicates that apatite only comprises a very small proportion. Therefore, the primary of apatite role in controlling the whole-rock REE budget can be excluded. As a result, the REE contents of the dyke come from the hydrothermal REE-bearing minerals. That is, hydrothermal fluids dominate the REE mineralization of the dyke. The fluids shall be the same as the post-magmatic fluids responsible for the formation of Type 2 apatite mentioned above.
Bulk rock REE contents of other carbonatite dykes in the ore district have also been reported in previous studies, and the data vary greatly between different types of dykes, different dykes of the same type, or different zones of the same dyke. For example, REE contents of rock samples from different Dol-dykes range from~100 ppm [21] to >15,000 ppm [8]. As field observations show that a Cal-dyke crosscut, a Dol-dyke, and a Cal-dyke have higher REE contents and (La/Yb) N ratios, a magnesio-ferro-carbonatite to calcio-carbonatite evolution sequence is thus established [21,28,29]. Liu et al. [25] proposed that calcio-carbonatites represented residual melts of evolved carbonatite magma and are rich in LREE, Sr, and Ba. However, this magmatic evolution process is not consistent with the evolution trend of most carbonatites in the world [49] and also cannot explain the heterogeneous REE mineralization of the carbonatite dykes in the ore district. Alternatively, our study results highlight that late-stage fluids expelled from the carbonatitic magmas dominate the REE mineralization of most dykes. The fluids interacted with dykes and wall rocks, causing heterogeneous REE mineralization of the dykes, transformation of some primary magmatic minerals, and different degrees of fenitization of wall rocks.

Implication for the Formation of the Bayan Obo Deposit
There are fine-grained and coarse-grained dolomite rocks in the Bayan Obo REE deposit according to the sizes of constituent minerals. The fine-grained dolomite rocks host more intense REE mineralization compared to the coarse-grained rocks [7,17]. Due to experiencing limited metasomatism and deformation, the coarse-grained dolomite rock is believed to be closer to the protolith before mineralization [25,26]. The total REE contents in coarse-grained dolomite mineral grains (Figure 7) are far lower than those of the whole rock (Figure 6b), similar to the case in the Dol-dyke mentioned above. Theoretically, carbonate minerals are supposed to have high capacity to incorporate REE for the substitution of Ca by REEs. However, if REEs in carbonatitic magmas can eventually form ore deposit, then REEs should be concentrated in the evolved melts or fluids rather than be dispersed in early formed carbonate minerals. Xu et al. [50] studied the Miaoya carbonatites and found that REE contents of calcite are only 100-260 ppm, far lower than those of the whole rocks. However, precisely because of the crystallization of these calcites, the REEs were able to concentrate in the residual melts and finally formed Miaoya REE ore deposit. Similarly, in the Bayan Obo ore district, dolomite grains from the Dol-dyke and ore-hosting dolomites contain low REE contents, indicating that early crystalized dolomite minerals consume little REE during carbonatitic melt evolution. Petrographic studies on the Dol-dyke in this study and ore-hosting dolomites in previous studies [25,44] reveal that REE-bearing minerals mainly occurred along dolomite grain boundaries or as late-stage veins in the dolomite rocks, implying that REE mineralization occurred later than the formation of dolomite minerals and was controlled by hydrothermal fluids. Therefore, it is highly possible that fluids derived from the evolved carbonatitic magmas controlled the REE mineralization of the deposit. The role of post-magmatic hydrothermal overprint in REE mineralization of carbonatites has also been proved and highlighted by many research studies worldwide [43,51,52]. Considering that the physicochemical conditions of REE mineralization were poorly constrained in the Bayan Obo deposit, it is important to conduct further studies on related fluid inclusions.
The ages of the deposit and carbonatite dykes have been dated by various testing methods, and the results cluster in periods at 1.4-1.3 Ga and 0.5-0.4 Ga [3,5,11,23]. Campbell et al. [53] reported SHRIMP 232 Th-208 Pb ages for a single zircon from the East orebody of 1325 ± 60 Ma in the zircon core and 455.6 ± 28 Ma at the rim. All these studies concluded that the deposit originally formed in the Mesoproterozoic and was subsequently influenced by hydrothermal fluids. Song et al. [27] found that in situ Th-Pb ages for monazite from the deposit range from 361Ma to 913 Ma and interpreted these ages as results of subsequent thermal perturbations. In addition, in situ analysis of other minerals indicate that the deposit underwent significant modification events after its formation, as evidenced by the addition of radiogenic Sr and Pb in dolomite and apatite [26,27] and abrupt changes in apatite compositions [44,54]. In this study, the presence of the Type 3 apatite in the dyke directly proves the occurrence of such a distinct hydrothermal activity in the Bayan Obo area. Although it is hard to tell the exact time of the disturbance, the fluids, according to the forming mechanism for Type 3 apatite, contain low contents of REE, Na, and Si contents. That is to say, the most outstanding disturbing fluids that occurred in the area probably provided no extra REEs to the Bayan Obo carbonatites.

Conclusions
(1) There are three types of apatite in the Dol-dyke. Type 1 apatite represents the primary apatite, which survived subsequent modification. After crystallization, the primary apatite in the dyke interacted with fluids of two separate stages: the first stage of metasomatism by post-magmatic fluids with fairly high LREE, Na, and Sr contents derived from evolved carbonatitic magmas, followed by a second stage of fluid infiltration that caused leaching of LREE and Sr from the primary apatite.
(2) The REE contents of the Dol-dyke in this study were controlled by hydrothermal fluids expelled from the carbonatitic magmas as demonstrated by widespread REE-bearing hydrothermal minerals as well as fluid-induced REE-rich Type 2 apatite.
(3) The studying results on the Dol-dyke of this study suggest that post-magmatic fluids derived from the evolved carbonatite melts played a key role in the REE mineralization of the Bayan Obo carbonatites, and a significant fluid disturbance occurred but provided no extra REEs for the carbonatites.