Magmatic-Hydrothermal Processes Associated with Rare Earth Element Enrichment in the Kangankunde Carbonatite Complex, Malawi

Carbonatites undergo various magmatic-hydrothermal processes during their evolution that are important for the enrichment of rare earth elements (REE). This geochemical, petrographic, and multi-isotope study on the Kangankunde carbonatite, the largest light REE resource in the Chilwa Alkaline Province in Malawi, clarifies the critical stages of REE mineralization in this deposit. The δ56Fe values of most of the carbonatite lies within the magmatic field despite variations in the proportions of monazite, ankerite, and ferroan dolomite. Exsolution of a hydrothermal fluid from the carbonatite melts is evident based on the higher δ56Fe of the fenites, as well as the textural and compositional zoning in monazite. Field and petrographic observations, combined with geochemical data (REE patterns, and Fe, C, and O isotopes), suggest that the key stage of REE mineralization in the Kangankunde carbonatite was the late magmatic stage with an influence of carbothermal fluids i.e. magmatic–hydrothermal stage, when large (~200 µm), well-developed monazite crystals grew. The C and O isotope compositions of the carbonatite suggest a post-magmatic alteration by hydrothermal fluids, probably after the main REE mineralization stage, as the alteration occurs throughout the carbonatite but particularly in the dark carbonatites.


Introduction
Rare earth elements (REE) are critical and strategic metals [1] and are becoming increasingly important for industrial use [2], thus warranting further exploration. Among the main types of REE deposits, carbonatites are important exploration targets because of their enrichment in these elements as well as their size and, generally, amenable mineralogy for extraction [3,4]. Although carbonatites exhibit notable enrichments in REE [5], the concentrations vary significantly, depending on the evolutional processes that they underwent, thereby allowing some carbonatites to be (potential) REE resources (e.g., Barra do Itapirapuã, Brazil [6]; the Mianning-Dechang and Qinling belts, China [7]; Kangankunde, Tundulu, and Songwe, Malawi [8,9]).
Processes that govern the enrichment of REE in carbonatites include fractional crystallization of a carbonatitic melt [10], enrichment by orthomagmatic fluids during late magmatic stages i.e. the magmatic-hydrothermal phase [5], dissolution and re-crystallization of primary to secondary carbonatitic minerals, resulting in the REE being concentrated in the latter [8], and supergene enrichment in weathered zones [11]. Although some studies report that magmatic processes play significant complex and a northern and southern knoll. There is a distinct zonation with regards to the lithologies of the main complex. The central core lithology which consists of carbonatite agglomerate is cut by carbonatite dykes and surrounded by a ring of feldspathic breccia and agglomerate [16]. The first intrusion stage comprises apatite-dolomite characteristics and followed by the intrusion of dolomitic to ferroan-dolomitic carbonatite [5]. The apatite-dolomite carbonatite was named beforehand by Garson [15], but the samples that exhibit similar characteristics in our study are named apatite-rich carbonatites because they were not collected from the main apatite body, but most likely from dykes that were too small to map. On the other hand, the terminology from Broom-Fendley et al. [5] is retained in this work, as the samples display similar physical characteristics that were used to name the samples in their study. The Kangankunde Carbonatite Complex is similar to other carbonatites in that it comprises a carbonatite body surrounded by a fenite aureole  [5,15]). Inset shows the location of Kangankunde in Malawi, Africa.

Local Geology
The main geological units of the Kangankunde Carbonatite Complex comprise Mn-and REErich carbonatites, dolomite carbonatites, apatite-rich carbonatites, carbonatite agglomerate, quartzrich rocks, and a fenite aureole that surrounds the main carbonatite body. The REE-rich carbonatites are concentrated in the center of the complex, with others occurring irregularly in veins and dykes (commonly as dolomitic and ankeritic carbonatites) that intrude different units of the carbonatite body.  [5,15]). Inset shows the location of Kangankunde in Malawi, Africa.

Local Geology
The main geological units of the Kangankunde Carbonatite Complex comprise Mn-and REE-rich carbonatites, dolomite carbonatites, apatite-rich carbonatites, carbonatite agglomerate, quartz-rich rocks, and a fenite aureole that surrounds the main carbonatite body. The REE-rich carbonatites are concentrated in the center of the complex, with others occurring irregularly in veins and dykes (commonly as dolomitic and ankeritic carbonatites) that intrude different units of the carbonatite body.
Throughout the complex, outcrops of carbonatite are commonly cut by veins that formed from late (carbothermal) intrusions (Figure 2a,b). Faults occur both regionally and locally in some of the carbonatites of the Chilwa Alkaline Complex. These faults are different from the local faults that developed at Kangankunde, which commonly strike NW and are absent in other carbonatite bodies in the Chilwa Alkaline Province. Well-developed monazite veins of a few cm in width are ubiquitous in the Kangankunde complex and cut carbonatite and fenite breccias (Figure 2c). Carbonatite agglomerates with distinct bulbous textures that resulted from weathering of the carbonate were observed in some lithologies (Figure 2d).
Minerals 2019, 7, x FOR PEER REVIEW 4 of 22 Throughout the complex, outcrops of carbonatite are commonly cut by veins that formed from late (carbothermal) intrusions (Figure 2a,b). Faults occur both regionally and locally in some of the carbonatites of the Chilwa Alkaline Complex. These faults are different from the local faults that developed at Kangankunde, which commonly strike NW and are absent in other carbonatite bodies in the Chilwa Alkaline Province. Well-developed monazite veins of a few cm in width are ubiquitous in the Kangankunde complex and cut carbonatite and fenite breccias (Figure 2c). Carbonatite agglomerates with distinct bulbous textures that resulted from weathering of the carbonate were observed in some lithologies (Figure 2d).

Methodology
Samples were acquired from the various lithological units in the Kangankunde Carbonatite Complex to obtain representative samples of mineralized and barren rocks that exhibit the different types of alteration. An extensive sampling program was carried out on the western side of the complex that is prominently fractured and hosts abundant dolomitic intrusions. Although the center of the complex is the main host for highly mineralized carbonatites, the presence of well-developed unidirectional monazite crystals on the west of the complex raised the need of the specific sampling strategy to clarify if the mineralization processes are similar in the various sections of the complex.
Monazite, which is the main host for the REE, imparts a green color to the carbonatites, allowing mineralized rock to be distinguished from barren rock. Samples were collected and classified into light ( Figure 3a) and dark (Figure 3b) carbonatites based on their physical appearance [5], apatite-rich carbonatites (Figure 3c), and fenites ( Figure 3d). Representative samples were crushed to <75 µm powder for further analysis.

Methodology
Samples were acquired from the various lithological units in the Kangankunde Carbonatite Complex to obtain representative samples of mineralized and barren rocks that exhibit the different types of alteration. An extensive sampling program was carried out on the western side of the complex that is prominently fractured and hosts abundant dolomitic intrusions. Although the center of the complex is the main host for highly mineralized carbonatites, the presence of well-developed unidirectional monazite crystals on the west of the complex raised the need of the specific sampling strategy to clarify if the mineralization processes are similar in the various sections of the complex.
Monazite, which is the main host for the REE, imparts a green color to the carbonatites, allowing mineralized rock to be distinguished from barren rock. Samples were collected and classified into light ( Figure 3a) and dark (Figure 3b) carbonatites based on their physical appearance [5], apatite-rich carbonatites (Figure 3c), and fenites ( Figure 3d). Representative samples were crushed to <75 µm powder for further analysis. Polished thin sections were examined using an Olympus BX60 optical microscope and a JEOL JSM-6510 scanning electron microscope (SEM). The chemical composition of minerals was semiquantitatively characterized using an energy-dispersive X-ray spectrometer (EDS) operated at an accelerating voltage of 20 kV and beam diameter of ~5 µm. Quantitative mineral compositions were obtained from representative thin sections using a JEOL JXA-8530F electron probe microanalyzer (EPMA, JEOL JXA-8530F) at the Laboratory of Nano-micro Material Analysis, Hokkaido University, Sapporo, Japan. Analyses were obtained using four X-ray detectors and an accelerating voltage of 15 kV, beam current of 20 nA, and beam diameter of ~5 µm. Instrument calibration was achieved by ZAF correction using commercial natural standards of LaP5O14, CeP5O14, NdP5O14, SmP5O14, EuP5O14, GdP5O14, DyP5O14, ErP5O14, YbP5O14, YP5O14, (Zr, Y) O2, and CaSiO3 provided by JEOL. Th, Ho, and Pr concentrations were determined by semi-quantitative analysis. All possible overlapping peaks were checked in pre-qualitative analysis. Lα X-ray lines were used for most quantitative analyses except for Sm and Nd concentration, which were determined by Lβ lines to avoid overlapping with Nd and Ce lines, respectively. Qualitative line analysis was also performed to obtain the concentration variations in the chemically zoned monazite. In the line analysis, peak positions of detected X-ray lines were confirmed using the same commercial standards as quantitative analysis.
Whole-rock major-and trace-element (except REE) compositions were analyzed using a Spectris, MagiX PRO X-ray fluorescence (XRF) spectrometer with a Rh tube. A mixture of 0.4 g of sample powder and 4 g of Li2B4O7 was placed in a platinum crucible, heated at 1000 °C for 8 min in a TK-4100 bead sampler, and then cooled, resulting in fused glass that was used for whole-rock XRF analyses. To avoid erroneous measurements due to incomplete dissolution, the REE concentrations were quantified from the glass bead samples that were initially prepared for XRF, using an Agilent 7700xc inductively coupled plasma-mass spectrometer equipped with a New Wave Research NWR213 laser Polished thin sections were examined using an Olympus BX60 optical microscope and a JEOL JSM-6510 scanning electron microscope (SEM). The chemical composition of minerals was semi-quantitatively characterized using an energy-dispersive X-ray spectrometer (EDS) operated at an accelerating voltage of 20 kV and beam diameter of~5 µm. Quantitative mineral compositions were obtained from representative thin sections using a JEOL JXA-8530F electron probe microanalyzer (EPMA, JEOL JXA-8530F) at the Laboratory of Nano-micro Material Analysis, Hokkaido University, Sapporo, Japan. Analyses were obtained using four X-ray detectors and an accelerating voltage of 15 kV, beam current of 20 nA, and beam diameter of~5 µm. Instrument calibration was achieved by ZAF correction using commercial natural standards of LaP 5  , and CaSiO 3 provided by JEOL. Th, Ho, and Pr concentrations were determined by semi-quantitative analysis. All possible overlapping peaks were checked in pre-qualitative analysis. Lα X-ray lines were used for most quantitative analyses except for Sm and Nd concentration, which were determined by Lβ lines to avoid overlapping with Nd and Ce lines, respectively. Qualitative line analysis was also performed to obtain the concentration variations in the chemically zoned monazite. In the line analysis, peak positions of detected X-ray lines were confirmed using the same commercial standards as quantitative analysis.
Whole-rock major-and trace-element (except REE) compositions were analyzed using a Spectris, MagiX PRO X-ray fluorescence (XRF) spectrometer with a Rh tube. A mixture of 0.4 g of sample powder and 4 g of Li 2 B 4 O 7 was placed in a platinum crucible, heated at 1000 • C for 8 min in a TK-4100 bead sampler, and then cooled, resulting in fused glass that was used for whole-rock XRF analyses. To avoid erroneous measurements due to incomplete dissolution, the REE concentrations were quantified from the glass bead samples that were initially prepared for XRF, using an Agilent 7700xc inductively coupled plasma-mass spectrometer equipped with a New Wave Research NWR213 laser ablation sampling system (LA-ICP-MS; Table S1) at the National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan.
Kangankunde carbonatites have minimal non-carbonate minerals, thereby the C isotopic compositions were taken to represent those of the carbonates, despite the homogenization of the different carbonate generations that may occur. C and O isotope ratios were quantified using a Thermo Scientific Delta V Plus isotope ratio mass spectrometer (IRMS). Powder samples were weighed, with respect to the amount of carbonate carbon contained, in tin containers to minimize sample loss due to electrostatic forces and transferred into 12 mL round-bottomed borosilicate vials [31] that were used to allow uniform flow of the gas to the IRMS. Carbon dioxide was extracted from the whole-rock powders by reaction with 100% phosphoric acid for 1 hour at 70 • C. The oxygen isotope fractionation factor between dolomite and CO 2 gas at 70 • C [32] was used to obtain the δ 18 O value of the carbonate samples. The JLs-1 limestone reference material was used as the laboratory standard (δ 13 C VPDB = 1.92% and δ 18 O SMOW = 31.09% ). Reproducibility of C and O isotope measurements, based on six analyses of this reference material, was better than 0.1% (2σ) for both δ 13 C and δ 18 O values.
Fe isotopes were measured using a Thermo Scientific Neptune Plus multi-collector-ICP-MS (MC-ICP-MS) at the Research Institute for Humanity and Nature, Kyoto, Japan. The procedure for sample digestion using acids was done following the method described as method C by Yokoyama et al. [33], where HCl, HNO 3 , HF, and HClO 4 were used for digesting. Purification of the iron solution was completed by chromatographic separation using columns compacted with anion-exchange resin (AG1-X8 200-400 mesh, Bio-Rad) based on the method of Ito et al. [34]. Solution samples were loaded into the columns with 1 mL of 8 M HCl, and the matrix was removed by 5 mL 8 M HCl and 5 mL of 3 M HCl. Purified Fe eluates were collected with 4 mL of 0.4 M HCl. The purification process was repeated, to improve the purity and multiple passes did not result in Fe isotope fractionations [24]. Finally, samples were redissolved and diluted in 1% HNO 3 to obtain 5 mL of 0.5 ppm Fe solution.
The Fe isotopic compositions were obtained in a middle resolution mode (M/∆M = 8000-9000) and the instrumental mass bias drift was corrected using the standard-sample-standard bracketing method [23]. Further interferences of 54 Cr + on 54 Fe + and 58 Ni + on 58 Fe + were corrected using the isotopic abundances ratios of 54 Cr/ 52 Cr = 0.0282 and 58 Ni/ 60 Ni = 2.616. The Fe isotopic compositions were assessed using standards and repeated analyses. The isotopic compositions are reported in standard δ-notation (per-mil, % ) relative to the standard reference material IRMM-014: The reproducibility of the Fe isotopic measurements, based on 9 measurements of the 0.5 ppm Fe standard solution, was typically better than 0.10% (2σ).

Whole-Rock Geochemistry
The major-and trace-element compositions of the carbonatites are summarized in Table A1. Based on the carbonatite classification diagram of Grittins and Harmer [35], which uses the molar proportions of CaO, MgO, and (FeO + MnO), the Kangankunde carbonatite are classified as magnesiocarbonatite and ferrocarbonatite (Figure 4). Although a larger group of the carbonatites are compositionally similar to dolomitic carbonatites, the presence of variable amounts of Fe results in a compositional trend towards ankeritic carbonatites. Similar trends were observed for most Kangankunde carbonatite data studied by Broom-Fendley et al. [5]. However, the extension towards Fe-rich end-members is prominent in the samples for this study. The carbonatites from the previous study were classified as mainly REE-rich carbonatites. The apatite-rich carbonatites are compositionally similar to the light and dark carbonatites, but the presence of Ca in the apatite-rich samples causes them to shift slightly towards the CaO region.   [35]). Arrows indicate the trends that occur with increasing contents of the specified minerals. Chemical compositoins of the Kangankunde carbonatites from a previous study [5] are plotted for comparison.
Chondrite-normalized plot of the Kangankunde carbonatite is presented in Figure 5. The Kangankunde carbonatite is generally strongly enriched in LREE ( Figure 5), except for the apatiterich carbonatites, which have shallower slopes on chondrite-normalized REE diagrams due to the higher contents of heavy REE (HREE; Figure 5). In addition, minimal Eu anomalies were observed but none of the carbonatites exhibit Ce anomalies in their REE patterns ( Figure 5). While the light carbonatites are less abundant than the dark carbonatites, some of them generally have the highest concentrations of REE. The dark carbonatites are the most abundant throughout the complex and are variably enriched in REE. The apatite-rich carbonatites generally have lower LREE concentrations but have HREE concentrations that are slightly higher than the light and dark carbonatites ( Figure 5).  [35]). Arrows indicate the trends that occur with increasing contents of the specified minerals. Chemical compositoins of the Kangankunde carbonatites from a previous study [5] are plotted for comparison.
Chondrite-normalized plot of the Kangankunde carbonatite is presented in Figure 5. The Kangankunde carbonatite is generally strongly enriched in LREE ( Figure 5), except for the apatite-rich carbonatites, which have shallower slopes on chondrite-normalized REE diagrams due to the higher contents of heavy REE (HREE; Figure 5). In addition, minimal Eu anomalies were observed but none of the carbonatites exhibit Ce anomalies in their REE patterns ( Figure 5). While the light carbonatites are less abundant than the dark carbonatites, some of them generally have the highest concentrations of REE. The dark carbonatites are the most abundant throughout the complex and are variably enriched in REE. The apatite-rich carbonatites generally have lower LREE concentrations but have HREE concentrations that are slightly higher than the light and dark carbonatites ( Figure 5).

Petrography
The Kangankunde Carbonatite Complex consists primarily of ferroan dolomitic and ankeritic carbonatites, the former are subdivided based on color into light and dark varieties. The carbonatites, which comprise ~70% ferroan dolomite and ankerite, range from pristine to highly altered iron oxides ( Figure 6). In the dark carbonatites, the carbonate matrix is mainly ankerite or ferroan dolomite, whereas dolomite is the major carbonate mineral in the light carbonatites. Two types of dark carbonatite are present in the complex. Although both types have been strongly altered into iron

Petrography
The Kangankunde Carbonatite Complex consists primarily of ferroan dolomitic and ankeritic carbonatites, the former are subdivided based on color into light and dark varieties. The carbonatites, which comprise~70% ferroan dolomite and ankerite, range from pristine to highly altered iron oxides ( Figure 6). In the dark carbonatites, the carbonate matrix is mainly ankerite or ferroan dolomite, whereas dolomite is the major carbonate mineral in the light carbonatites. Two types of dark carbonatite are present in the complex. Although both types have been strongly altered into iron oxides, they differ in that one type does not contain monazite or REE mineralization (Figure 6a), whereas the other type does (Figure 6b). Furthermore, the light carbonatites (Figure 6c) comprise mainly euhedral and unaltered dolomite and are comparatively less abundant throughout the complex. In contrast to the carbonate in the light and dark carbonatites, that in apatite-rich carbonatites has been replaced by Fe-Ti oxides in a fluidal pattern (Figure 6d). Iron oxides are present with varying contents of Fe, with the highly altered carbonatites having higher proportions of iron oxides. The iron oxides comprise hematite, goethite, and an unidentified Fe-Mn-Ba-bearing oxide. Calcite is absent in most of the Kangankunde carbonatites, with only minor amounts occurring in the apatite-rich carbonatites. A few other Kangankunde carbonatites that were observed have similar characteristics to the dark carbonatites but are not discussed further as they are categorized as an independent group given their higher proportions of quartz (~26.6%), likely the quartz-rich rocks as previously defined [15].  The main REE-bearing minerals in the Kangankunde Carbonatite Complex are monazite, bastnaesite, and synchysite. Monazite, which is the most abundant REE-bearing mineral, is green in hand specimen and exhibits a greenish brown color under plane-polarized light. Two varieties of monazite are present. One variant comprises large (~200 µm) well-developed crystals that typically occur along the boundaries of dolomite and is commonly observed with abundant iron oxides ( Figure  7a). These crystals vary from porous in the core to non-porous in the rim (Figure 7b). The second variant comprises small polycrystalline aggregates of monazite that rarely occur with iron oxides in hand specimen and exhibits a greenish brown color under plane-polarized light. Two varieties of monazite are present. One variant comprises large (~200 µm) well-developed crystals that typically occur along the boundaries of dolomite and is commonly observed with abundant iron oxides (Figure 7a). These crystals vary from porous in the core to non-porous in the rim (Figure 7b). The second variant comprises small polycrystalline aggregates of monazite that rarely occur with iron oxides (Figure 7c,d). Bastnaesite and synchysite are less common than monazite and occur as thin needle-like laths (Figure 7e,f)  Strontianite and barite are typically associated with monazite. They generally envelop the monazite or occur along crystal boundaries. Strontianite is typically anhedral and contains trails of fluid inclusions whereas barite typically occurs as euhedral laths ( Figure S2). Other minerals in the Kangankunde carbonatites include apatite, quartz, and iron oxides. (Fluor)apatite is most abundant in the apatite-rich carbonatites ( Figure S2), but also occurs in minor proportions in the other types of carbonatite. It typically occurs as anhedral ovoid or euhedral hexagonal crystals with minimal zoning. Altered apatite occurs as radiating needles that typically occur in a dolomite matrix. Furthermore, the large monazite crystals display chemical zoning as shown in the backscattered image in Figure 8 and Table A2. The rims are slightly enriched in HREE (Nd and Sm) compared to the cores. The distribution of LREE is quite homogeneous within the crystal but in minor cases shows a depletion of La and Ce in the rims. Although the texture of the monazite cores can either be porous or non-porous, they are compositionally similar in either case.

Carbon and Oxygen Isotopes
The C and O isotopic compositions of the Kangankunde carbonatites are given in Table A3 and plotted in Figure 9. The δ 13 C values of the carbonatites (-3.98‰ to 0.10‰) show a smaller variation compared with the δ 18 [21]. Although the data points plotted from a previous study on the Kangankunde carbonatites [5] spread through the plot, they have slightly lower δ 18 O values in comparison to our sample set. Mineral separates were analyzed for the previous study and the carbonates from the strontianite-rich rock and the beforsite (apatite-rich) plot very close to the PIC field. Figure 9. Whole-rock δ 13 C and δ 18 O compositions of the Kangankunde carbonatites compared with those in a pvevious study [5]. Compositional fields are shown for primary igneous carbonatites [29]. Strontianite and barite are typically associated with monazite. They generally envelop the monazite or occur along crystal boundaries. Strontianite is typically anhedral and contains trails of fluid inclusions whereas barite typically occurs as euhedral laths ( Figure S2). Other minerals in the Kangankunde carbonatites include apatite, quartz, and iron oxides. (Fluor)apatite is most abundant in the apatite-rich carbonatites ( Figure S2), but also occurs in minor proportions in the other types of carbonatite. It typically occurs as anhedral ovoid or euhedral hexagonal crystals with minimal zoning. Altered apatite occurs as radiating needles that typically occur in a dolomite matrix.

Carbon and Oxygen Isotopes
The C and O isotopic compositions of the Kangankunde carbonatites are given in Table A3 and plotted in Figure 9. The δ 13 C values of the carbonatites (−3.98% to 0.10% ) show a smaller variation compared with the δ 18 O values (12.57% to 28.94% ). The large range in δ 18 O values can be related to the type of carbonatite, with light carbonatites having low δ 18 O, dark carbonatites having high δ 18 O, and apatite-rich carbonatites having low to high δ 18 O values. Furthermore, the light carbonatites form an end member close to the Primary Igneous Carbonatites (PIC) field representing the δ 13 C and δ 18 O value ranges for primary, mantle derived carbonatites [21]. Although the data points plotted from a previous study on the Kangankunde carbonatites [5] spread through the plot, they have slightly lower δ 18 O values in comparison to our sample set. Mineral separates were analyzed for the previous study and the carbonates from the strontianite-rich rock and the beforsite (apatite-rich) plot very close to the PIC field.

Iron Isotopes
The δ 56 Fe values for the Kangankunde carbonatites range from −0.38% to −0.25% , with fenite having a value of −0.18% (Table A4). These values are similar to the average δ 56 Fe value for magmatic carbonatites (−0.30% ± 0.05% ) [23]. No correlation exists between the δ 56 Fe values and Fe contents of the samples. and apatite-rich carbonatites having low to high δ 18 O values. Furthermore, the light carbonatites form an end member close to the Primary Igneous Carbonatites (PIC) field representing the δ 13 C and δ 18 O value ranges for primary, mantle derived carbonatites [21]. Although the data points plotted from a previous study on the Kangankunde carbonatites [5] spread through the plot, they have slightly lower δ 18 O values in comparison to our sample set. Mineral separates were analyzed for the previous study and the carbonates from the strontianite-rich rock and the beforsite (apatite-rich) plot very close to the PIC field. Figure 9. Whole-rock δ 13 C and δ 18 O compositions of the Kangankunde carbonatites compared with those in a pvevious study [5]. Compositional fields are shown for primary igneous carbonatites [29].  [5]. Compositional fields are shown for primary igneous carbonatites [29].

Evolution of the Carbonatites
The Kangankunde carbonatites exhibit features that are typical of magmatic, magmatichydrothermal, and post magmatic processes through their evolution. The minimal alteration in some minerals such as the carbonate in the light carbonatite (Figures 6c and 7d), the small windows of δ 13 C and δ 56 Fe isotopic compositions and the geochemistry of some of the carbonatites support a magmatic stage involved in their formation, which corresponds with the interpretation made by Broom-Fendley et al. [5] based on their Sr isotopes. On the other hand, the significantly altered mineralogy in dark and apatite-rich carbonatites, the elevated δ 18 O isotopic compositions and enrichment of HREE in monazite cores and apatite-rich samples are indicative of the late to post magmatic events through the carbonatites' evolution, also partly reported by Raymond [16] and Broom-Fendley et al. [5].
The carbonatites presented in this study are from various lithologies and focusing on the western section of the complex, which also display high REE concentrations (>2 wt. %). Using the carbonatite classification (Figure 4), these carbonatites spread from magnesiocarbonatite to ferrocarbonatite fields, indicating that they have higher Fe concentrations than most of the carbonatites from the complex that were reported previously [5]. Despite the higher concentrations observed in this study, the light carbonatites have comparatively lower Fe concentrations than the dark and apatite-rich carbonatites. This has been attributed to the breakdown of the ferroan dolomite resulting in darker mineral grains [37] in the dark carbonatite, which supports the occurrence of alteration processes.
However, the Fe isotopic compositions obtained from the carbonatites occur within a very small window (−0.38% to −0.26% , Figure 10), which is within the Fe isotopic range defined for magmatic carbonatites (−0.30% ± 0.05% ) [23,24]. While igneous rocks have demonstrated to have homogenous Fe isotopic compositions [38], high temperature alteration and/or fluid fractionation may result in the fractionation of Fe isotopic values [26,39], hence minimal fractionations would account for significant processes. The evolutional path ( Figure S3) reported for the carbonatite affects the behavior of the REE in the carbonatite, hence the need to clarify the roles of the respective processes on REE enrichment. that scarce mineralization occurs during the primary magmatic stage. Accordingly, few REE mineralization events of the Kangankunde carbonatites are magmatic in origin.

Magmatic-Hydrothermal Stages Associated with REE Enrichment
Due to the high susceptibility of carbonatites to metasomatization and alteration [5], magmatic fluids are released through the process, which may result in REE enrichment during alteration. Among others, a close observation on the monazite textural variations observed in the Kangankunde carbonatite provides insight on the mineralization events. The textures exhibited by the monazite (Figure 7) depend on whether they occur in the light or dark carbonatites. Monazite in the light carbonatites is typically small in size and occurs as polycrystalline aggregates, whereas that in the dark carbonatites typically occurs as relatively large and well-developed, hexagonal crystals. This textural variation suggests that the monazites have undergone various stages (e.g., hydrothermal) other than the primary magmatic processes.
The small aggregated textures exhibited by monazite in the light carbonatites ( Figure 7) may have resulted from fractional crystallization during the early magmatic stages, as the carbonatites solely preserves magmatic signatures from petrography and C isotopic compositions, despite occurring outside the PIC field. However, the study area contains late magmatic intrusions ( Figure  1) that indicate intricate occurrences during the evolution of the carbonatite which explains the monazite textures in the dark and apatite-rich carbonatites. The large (~200 µm) monazite crystals in the dark carbonatites, typically occur along the grain boundaries of dolomite, suggesting either that they precipitated from a later fluid or crystallized concurrently with the dolomite. The minor Eu anomalies also suggest formation of monazite in a metamorphic environment [45]. Furthermore, they exhibit homogenous textures and chemical zoning, features typical of ortho-to late-magmatic stages [10].
Monazite can form in metamorphic environments [45][46][47], as pseudomorphs of other minerals (e.g., burbankite) [5], or as an alteration product during the evolution of minerals such as apatite [48]. The typical crystallization temperature of monazite in a metasomatic environment is ~600 °C [49], which would allow for the formation of well-developed crystals similar to those observed in the Kangankunde dark carbonatites. Broom-Fendley et al. [5] suggested that the monazite crystals in the Kangankunde carbonatites are pseudomorphs that formed from the breakdown of burbankite immediately after its formation, an observation that was confirmed from the clear hexagonal shape of the large the monazite (Figure 8). In addition to the hexagonal shapes, the large monazite crystals

REE Mineralization in the Magmatic Stage
TREE concentrations in the light and dark carbonatites range from 2.53 to 23.71 wt. % and 1.41 to 16.65 wt. %, respectively. Given the magmatic isotopic signatures and minimally altered dolomite of the light carbonatites, the fact that they host significant concentrations of REE suggests that at least some of the REE mineralization occurred during the early magmatic stage despite that REE mineralization in most carbonatites is thought to result from the effects of (carbo-) hydrothermal fluids [3,14,40]. The presence of both apatite and monazite in the Kangankunde carbonatites suggests that the initial magma had high concentrations of P. In addition, the occurrence of REE mineralization in the magmatic stage indicates that the initial magma also had high concentrations of REE. Carbonatite associated REE deposits have been estimated to form from a parent magma that is extremely rich in the REE, along with Sr and Ba [41], which may result in an enrichment of these elements through crustal processes (e.g., liquid immiscibility [40] and subsolidus remobilization [42]). Such a REE-enriched initial magma could have been generated by partial melting of refertilized mantle that had been metasomatized by CO 2 -rich fluids [43]. Although experimental studies (e.g., [43,44]) have demonstrated that REE mineralization can result from fractional crystallization of carbonatitic melts due to the incompatibility of the REE and other associated elements, it is typical that scarce mineralization occurs during the primary magmatic stage. Accordingly, few REE mineralization events of the Kangankunde carbonatites are magmatic in origin.

Magmatic-Hydrothermal Stages Associated with REE Enrichment
Due to the high susceptibility of carbonatites to metasomatization and alteration [5], magmatic fluids are released through the process, which may result in REE enrichment during alteration. Among others, a close observation on the monazite textural variations observed in the Kangankunde carbonatite provides insight on the mineralization events. The textures exhibited by the monazite (Figure 7) depend on whether they occur in the light or dark carbonatites. Monazite in the light carbonatites is typically small in size and occurs as polycrystalline aggregates, whereas that in the dark carbonatites typically occurs as relatively large and well-developed, hexagonal crystals. This textural variation suggests that the monazites have undergone various stages (e.g., hydrothermal) other than the primary magmatic processes.
The small aggregated textures exhibited by monazite in the light carbonatites ( Figure 7) may have resulted from fractional crystallization during the early magmatic stages, as the carbonatites solely preserves magmatic signatures from petrography and C isotopic compositions, despite occurring outside the PIC field. However, the study area contains late magmatic intrusions (Figure 1) that indicate intricate occurrences during the evolution of the carbonatite which explains the monazite textures in the dark and apatite-rich carbonatites. The large (~200 µm) monazite crystals in the dark carbonatites, typically occur along the grain boundaries of dolomite, suggesting either that they precipitated from a later fluid or crystallized concurrently with the dolomite. The minor Eu anomalies also suggest formation of monazite in a metamorphic environment [45]. Furthermore, they exhibit homogenous textures and chemical zoning, features typical of ortho-to late-magmatic stages [10].
Monazite can form in metamorphic environments [45][46][47], as pseudomorphs of other minerals (e.g., burbankite) [5], or as an alteration product during the evolution of minerals such as apatite [48]. The typical crystallization temperature of monazite in a metasomatic environment is~600 • C [49], which would allow for the formation of well-developed crystals similar to those observed in the Kangankunde dark carbonatites. Broom-Fendley et al. [5] suggested that the monazite crystals in the Kangankunde carbonatites are pseudomorphs that formed from the breakdown of burbankite immediately after its formation, an observation that was confirmed from the clear hexagonal shape of the large the monazite (Figure 8). In addition to the hexagonal shapes, the large monazite crystals display distinct chemical zoning that was not observed in the small aggregates of monazite of the light carbonatites. Monazite crystal variation is known to result from various stages of (re)crystallization, which may cause compositional variations especially after interacting with a hydrothermal fluid [42]. Similar processes may have occurred in the Kangankunde large monazite. The chemical zoning in these crystals was then a result of the reaction with hydrothermal fluids. We therefore suggest that the large monazite crystals formed following an alteration from burbankite, and that compositionally zoned monazite was formed due to the involvement of a hydrothermal fluid at near-magmatic temperatures. This is supported by the occurrence of these monazites as large and well-developed crystals, and by their compositional and textural zoning (Figure 8).
The significance of late-magmatic processes to the generation of REE-enriched minerals has been reported [50]. In the early stages of carbonatite crystallization, fluids typically exsolve from the carbonatite melt due to their high-water contents [25]. These fluids are usually enriched in REE because there is, in general, a low mineral/melt partition coefficient for the REE [13]. In the case of a much later fluid, interaction with a residual magma that had high concentrations of REE may also result in enriched REE concentrations in the fluid phase [13]. Exsolution of a fluid phase rich in Fe 3+ , Cl, F, S, Ca, Mg, or other carbonatite constituents from the carbonatite magma has been stated [51] and, in the Kangankunde carbonatite, is supported by the heavy δ 56 Fe values in fenites relative to the carbonatites [23,24,52]. Furthermore, significant remobilization signatures are observed in outcrops of the Kangankunde carbonatites (Figure 2a,c) and are particularly prominent in the western region where unidirectional monazite veins are observed, which suggests the presence of a fluid. Studies of carbonatites from Brazil [53] and India [13] have indicated that the late magmatic stages are crucial for the enrichment of REE. Hence, the precipitation of monazite from such an REE-rich fluid could have generated the REE enriched monazites in the Kangankunde carbonatites.
During the early magmatic stages of carbonatite formation, carbothermal fluids enriched in ligands such as OH − , F − , and Cl − typically exsolve from the carbonatitic magma [17]. Interaction of these fluids with the REE-rich residual magma would generate complexes with the REE, thereby enhancing the formation, transportation, and deposition of large, REE-rich hydrothermal monazite (Figure 7b). Based on the textural and compositional variations observed within the monazite crystals of this study (Figure 8), monazite occurred in multiple stages characterized by differences in pressure, temperature, and fluid composition. These changing conditions would also have affected the uptake of REE by monazite [54]. The occurrence of these monazites along the crystal boundaries of dolomite (Figure 6b), along with the alteration of carbonates (dolomite and/or ankerite) by Fe oxides, further suggests that the large monazite crystals formed as a result of a fluid infusion and possibly remobilization through the earlier formed carbonatites [13].
Within a magmatic-hydrothermal system, it has been suggested that Fe isotopes are fractionated due to (a) exsolution and loss of an aqueous fluid [25,54,55], (b) fractional crystallization, which may also involve changes in redox conditions [56][57][58], and (c) the effects of thermal diffusion [28,[59][60][61]. These processes commonly occur in magmatic-hydrothermal systems during and after the formation of carbonatites and can be characterized using Fe isotopes [62]. In the Kangankunde carbonatites, Fe is hosted mainly by ankerite and ferroan dolomite, which make up >70% of the carbonatites. Therefore, the fractionation of Fe isotopes reflects the major processes that led to carbonatite formation and, more importantly, to REE mineralization. The concentration of Fe is also notably different between the light and dark carbonatites, with the latter generally having higher concentrations (Figure 4). This compositional difference suggests either that Fe was added to the carbonatite by fluids (alteration of ferroan dolomite by iron oxides in the dark carbonatites) [5], that the light and dark carbonatites crystallized from compositionally distinct magmas (i.e., the dark carbonatite crystallized from a magma that was enriched in Fe) [63], or that the Fe is a product of exsolution in the breakdown of ferroan dolomite. In either case, the concentration of Fe among the different carbonatites varies significantly ( Figure 3) and is an important indicator of their evolutional processes.
The δ 56 Fe values for the Kangankunde carbonatites ( Figure 10) are relatively homogenous (−0.38% to −0.26% ) and indicate only minor fractionation among the carbonatites. The Fe-rich carbonates (dolomite and ankerite) are observed mainly in the dark carbonatites where they are further being replaced by Fe oxides through weathering. Despite the various controls on Fe in the carbonatite, Fe isotopic compositions display minimal fractionations from each other, all within that defined for magmatic carbonatites. This indicates that the major stages responsible for the formation of the carbonatites would have been during the magmatic window.
The difference in isotopic composition of the fenites and carbonatites (Figure 10), however, suggests that partial oxidation and exsolution of an Fe 3+ -bearing fluid with heavier δ 56 Fe values also played an important role in the evolution of the carbonatite. Exsolution of fluids from the carbonatite melt is consistent with the abundance of large monazite crystals (Figure 7), given that fluid exsolution likely occurred during the magmatic-hydrothermal phase, resulting in further alteration. The presence of the light carbonatites, which strongly preserve signatures of magmatic origin (Figure 9), suggests that during the early phases of carbonatite evolution, several stages might have occurred. These phases may have resulted from different pulses of magma; e.g., an initial pulse that had relatively low Fe contents followed by a subsequent pulse that was relatively enriched in Fe and possibly accompanied by fluid that resulted in alteration. The monazite crystals in the dark carbonatites are well developed, which is common for monazites that crystallize during the magmatic stage [20]. Their large size is a result of the subsequent magma pulses that interacted with magmatic fluids, resulting in only minor δ 56 Fe fractionation in the carbonatites. Therefore, despite the significant mineralogical and compositional differences among the carbonatites, the REE-bearing carbonatites likely formed during the magmatic-hydrothermal stage. The formation of minor amounts of REE-bearing minerals in the magmatic stage was likely due to the initial carbonatitic magma being relatively enriched in REE.
Crystallization of other minerals during the early stages of carbonatite formation may have generated the minor negative shift of δ 56 Fe value in some of the samples [23], which likely occurred late in the evolution series. It is evident, however, that significant evolution processes, including REE mineralization, occurred during the magmatic-hydrothermal stages of carbonatite formation. Furthermore, the REE enrichment of the carbonatites on the western side of the complex was enhanced by the magmatic-hydrothermal processes, resulting in the dark carbonatites that are significantly altered in comparison to the light carbonatites.

Low-Temperature Hydrothermal Overprint
The magmatic δ 13 C and δ 18 O signatures of the carbonatites were modified by post-magmatic processes (Figure 9). Compared with δ 18 O, the δ 13 C values of the carbonatites are less variable (−0.03% to −4.21% ) and fall within the range of magmatic values but above the range of mantle values (δ 13 C = −7.0% to −5.0% ) [64], suggesting that the magmatic C isotope signature was preserved. The elevated δ 13 C values compared with those expected of mantle-derived melts was likely caused by near-surface alteration by fluids that were meteoric in origin, possibly after exhumation of the complex [8,65]. In contrast to the δ 13 C values, the δ 18 O values (13.61% to 28.61% ) are more variable and generally correlate with the physical appearance of the carbonatites, with the light carbonatites having lower values than the dark carbonatites. Fractionation models calculated by Ray and Ramesh [66] and Broom-Fendley et al. [5] suggest that trends towards high O isotopic compositions, similar to those observed in the Kangankunde carbonatites, can result from the alteration of carbonatites by low-temperature fluids. The temperatures predicted for these fluids (<150 to~250 • C) in a model presented by Broom-Fendley et al. [5] are consistent with the homogenization temperatures obtained from fluid inclusions that were observed in the dominant carbonate minerals in the Kangankunde carbonatites ( Figure S4) but seemed to have not been related to the main mineralization event due to their significantly low temperatures and salinities. This is indicative of an overprint by low-temperature fluids that modified the O isotopic composition of the carbonatites, probably after REE mineralization.

Conclusions
The magmatic and post-magmatic stages that were critical in the enrichment of REE in the Kangankunde carbonatites have been characterized by integrating field and petrographic observations with geochemical and isotopic data. The magmatic origin of the Kangankunde Carbonatite Complex is demonstrated by the presence of pegmatitic minerals and magmatic Fe isotopic signatures.
Post-magmatic alteration is demonstrated by complex outcrops, mineral textures and O isotopic signatures. The magmatic processes resulted in the formation of the volumetrically minor light carbonatites, which are variably enriched in REE. Interaction of the residual magma with a magmatic fluid augmented the crystallization of the hydrothermal fluid-related hexagonal, chemically zoned monazite that is abundant in the dark carbonatites, which are notably enriched in REE.
After solidification, the carbonatites were overprinted by low-temperature fluids that were likely meteoric in origin, resulting in the formation of low-temperature alteration signatures and low homogenization temperatures for fluid inclusions. Based on Fe, C, and O isotopes, the formation of the carbonatites and the REE mineralization occurred largely during the early to late magmatic stages, whereas the alteration modified the carbonatites but was not related to REE mineralization. This suggests that late magmatic fluids of the magmatic-hydrothermal stage were critical in the REE enrichment of the Kangankunde carbonatites.