Geology and U-Th-Pb Dating of the Gakara REE Deposit, Burundi

The Gakara Rare Earth Elements (REE) deposit is one of the world’s highest grade REE deposits, likely linked to a carbonatitic magmatic-hydrothermal activity. It is located near Lake Tanganyika in Burundi, along the western branch of the East African Rift. Field observations suggest that the mineralized veins formed in the upper crust. Previous structures inherited from the Kibaran orogeny may have been reused during the mineralizing event. The paragenetic sequence and the geochronological data show that the Gakara mineralization occurred in successive stages in a continuous hydrothermal history. The primary mineralization in bastnaesite was followed by an alteration stage into monazite. The U-Th-Pb ages obtained on bastnaesite (602 ± 7 Ma) and on monazite (589 ± 8 Ma) belong to the Pan-African cycle. The emplacement of the Gakara REE mineralization most likely took place during a pre-collisional event in the Pan-African belt, probably in an extensional context.


Introduction
During the recent years, the scientific community made an important effort in order to better understand the metallogenic conditions behind the emplacement of Rare Earth Elements (REE) mineralization. Alkaline intrusions and carbonatites and the associated hydrothermal activities are known to host the largest REE deposits [1]. Among the world-famous occurrences, the Gakara deposit (Burundi) is described as a typical example of hydrothermal deposits developed in association with carbonatite magmatism [2]. Nevertheless, little is known or has been published [3,4] about this deposit, especially when compared to the Bayan Obo deposit in China. Indeed, during the last 10 years, more than 100 articles have been published on this unique deposit [5]. Since September 2017, the Gakara deposit is mined by the "Rainbow Rare Earths" mining company with a targeting run rate of about 5000 tons per year. With in-situ grades in the range of 47-67% Total Rare Earth Oxide, Gakara is one of the world's richest rare earth deposits. The mining operations follow a period of intensive exploration, during which new outcrops have been exposed and reference samples have been collected.
In this work, we first report the field observations we made during two fieldwork campaigns, in October 2017 and February 2018, respectively. These field data allow us to define the Finally, the Cenozoic rifting ( Figure 1B) affected the whole area with a general uplift. Normal faults that accommodated this rifting are not clearly recognized in the Gakara area; however, some large blocks could have been tilted during these extensional tectonic phases.  [3,12,22,23]. (B) Geological map of the study area showing the occurrences of REE mineralization (Rainbow Burundi Mining data) and their host rocks; compiled data and redrawn from the 1/100,000 geological maps of Burundi: sheet Rumonge [24], sheet Bujumbura [25], sheet Bururi and sheet Gitega [26].

Host Rocks
Rocks are poorly exposed in the Gakara region, as they are often buried underneath a lateritic cover, and because of the presence of numerous agricultural crops and slope deposits. Some outcrops are nevertheless visible along new roads opened for mining purposes, as well as along pedestrian roads and river beds. All the visited sites are enclosed within gneisses. The gneisses are very heterogeneous in term of lithologies and structures. The most common gneiss facies corresponds to an orthogneiss with K-feldspar phenocrysts, biotite, quartz, with less abundant muscovite ( Figure 2A) likely deriving from S-type granites resulting from the partial melting of the surrounding paragneisses during the Kibaran orogeny. The other type of gneisses corresponds to FK-paragneisses well exposed at Kiyenzi. The original sedimentary bedding transposed by a former S1 foliation is folded with an associated S2 axial planar cleavage ( Figure 2B).
The deformation that affected these gneisses, visible at all scales, corresponds to a system of ductile shear bands. Sub-solidus and solid-state deformation have been observed. The existence of sub-solidus and hot deformation is demonstrated by the presence of: (i) typical C/S structures in the orthogneisses, the C planes being underlined by biotite ( Figure 2A); (ii) ductile shear zones where pegmatites were emplaced ( Figure 2C). Such structures argue for a syntectonic emplacement of the granitoids. Solid-state deformation is attested by the occurrence of foliated mylonites associated with an important grain size reduction. Both the ortho-and para-gneisses are intruded by numerous pegmatitic bodies ( Figure 2D,E). When the foliation plan (S1) is parallel to the exposed surface, some of these bodies appear as classical dykes and sills crosscutting the foliation ( Figure 2D). However, most of them are transposed into the foliation and strongly deformed ( Figure 2E). They are also folded and schistosed by a second foliation (S2) in the paragneiss ( Figure 2B). Internal deformation of pegmatites can be very strong parallel to the main foliation (S1) with biotite marking shear planes and a strong stretching of both quartz and feldspar. The pegmatites parallel to S0/S1 are folded and schistosed by S2; (E) pegmatite and host gneiss, the fold affects both the pegmatite and a previous foliation (main road to Gakara).

Types of Mineralization
Despite the lack of outcrops, several types of mineralization can be observed on the field. Here we present the gitology of the Gasagwe mine and the most important REE occurrences recognized and explored in the Gakara claim (see Figure 1B for mine and occurrences locations). It has to be noticed that all the visited pits and trenches underwent strong laterization and weathering (picture in Figure 3).
The Gasagwe ore deposit is the most important mining site, where production started in September 2017. There, a spectacular anastomosed network of REE veins trends N to NNE and dips gently westward (Figure 3). Many veins present strike and thickness variations and therefore display a curved aspect. As the result, the variable distribution of measures observed in stereogram ( Figure 3) is mostly due to that curved geometry rather than various sets of vein orientations. Most of the REE vein wall rocks correspond to former anisotropies within the basement. Indeed, a white or green fine foliated rock is frequently observed along the REE-vein roof wall ( Figure 4A); these foliated zones correspond to foliated pegmatites or mylonites ( Figure 4D). Some of these shear zones, lately reused by the REE-bearing fluids, correspond to very flat thrusts ( Figure 4B), likely Kibaran in age. Veins texture does not bear any ductile shear component and corresponds to brittle opening, with bastnaesite crystalizing in the resulting open spaces. "Monazitization" of primary bastnaesite occurred during a subsequent brittle increment responsible for mosaic-like breccia of former bastnaesite crystals. Many wall rock clasts (gneiss, pegmatites) are incorporated within the ore, a feature in a good agreement with the brecciation of the REE vein wall rocks by numerous ore veinlets ( Figure 4C), which corresponds to an irregular damage zone formed during the REE mineralizing event. The pattern of this damage zone is characteristic of hydrothermal breccia formed by hydraulic fracturing during REE-bearing fluid circulations. Finally, cm-to dm-long dark idiomorphic quartz crystals are frequently found between the hydraulic breccia and the massive ore. The long axes of these crystals appear to be randomly oriented between a normal to a parallel orientation with respect to the contact between the wall rock and the REE-bearing veins. These dark quartz crystals seem to be associated with the monazite alteration rather than with an early bastnaesite crystallization.  The Kiyenzi occurrence is capped by a hill made of hard and massive white aplitic rocks ( Figures 5  and 6D), the color being related to a pervasive albitization of the dykes. The albitized aplitic dyke walls localized a cataclastic shearing responsible for the formation of a foliated cataclasite affecting both the aplites and their paragneissic host ( Figure 6B,C). There is no evidence for a ductile deformation in the area. The REE mineralization corresponds to clasts of former veins embedded within cataclasite cemented by late iron oxides and quartz ( Figure 6E,F). Therefore, former REE-bearing veins, developed mainly along the walls of the aplitic dykes, have been fragmented and brecciated by post-REE cataclasis. This is well illustrated by the structural measurements ( Figure 6A) which are randomly distributed due to a late tectonic brecciation.
The Gashirwe occurrence is developed within white orthogneisses where mineralization is more frequently localized within and/or along pegmatites, white aplitic dykes and vertical fractures ( Figure 7). Here again, the REE-bearing veins reused older heterogeneities to propagate ( Figure 7D). The aplite dykes clustered dry jointing ( Figure 7B,C). The orientation and dips (sub-vertical) of these dry joints are comparable with those of the REE-bearing veins which argue for a cogenetism, at least at a local scale. Structural data obtained from the Gashirwe main outcrop ( Figure 7A) argue for a dominant N70E trending direction for the REE-bearing veins with a sub vertical dip. However, using the Rainbow Mining Burundi (RMB) database at the scale of the entire Gashirwe hill prospect, the contouring of the REE-bearing vein poles suggests that many veins are gently dipping with variable strikes ( Figure 7A).   The main feature of the Murambi REE-bearing veins cluster is that nearly all the veins are hosted in Kibaran pegmatites which are strongly deformed in some places. The pegmatites and REE-bearing veins are enclosed within an orthogneiss. These veins are gently dipping and contain REE-bearing minerals filling discontinuities and embedded within laminated quartz ribbons ( Figure 8A). Moreover, the contact between the REE-bearing veins and the host pegmatites is characterized by the presence of corrosion gulfs and the incorporation of micas and quartz from the pegmatites into the ore ( Figure 8B). Finally, near the Masenga hill, in the north-east part of the mineralized area, a small cliff along the pedestrian path exposes a highly strained biotite-orthogneiss cut by a meter-thick pegmatitic vein ( Figure 9A). Here again, this pegmatitic vein is affected by the regional, steeply-dipping Kibaran foliation. Immediately above the pegmatite, a 20-cm wide, low dipping dyke of isotropic microgranite cuts the foliation. At the base of the microgranite dyke, a lenticular REE-ore body occurs, parallel to the dyke. In the lenticular ore, the several-cm long bastnaesite crystals tend to grow at a high angle to the edge, leaving empty spaces between them.

Ore Sample Description and Typology
Samples were collected during the RMB exploration campaigns (2011-2014). Additional samples come from some alluvial deposits present in the rivers (Nyarubenga, Mugere, Karonge, Kirasa) crossing the mineralized zone ( Figure 1B). Representative samples named GOMV, KIV, NYARU, GASH, GASA, MASE and MURA come, respectively, from the following localities: Gomvyi, Kivungwe, Nyarubenga, Gashirwe, Gasagwe, Masenga and Murambi (see locations Figure 2). Depending on the degree of alteration, different types of ore can be found: from massive pure bastnaesite ore to an almost pure monazite ore, and in-between, ores with various proportions of bastnaesite and monazite. Sample GOMV ( Figure 10A) shows a bastnaesite vein associated with a quartz vein. This sample represents the first stage of mineralization where monazite is absent. Sample KIV ( Figure 10B) exhibits weakly-altered crystals of bastnaesite with spaces in between partly filled with quartz. Primary textures corresponding to large euhedral bastnaesite laths intercalated with polyhedral cavities filled by quartz ( Figure 10A,B) resemble a "pegmatite-like" texture. Sample NYARU ( Figure 10C) represents a breccia ore with angular clasts of bastnaesite that are randomly oriented in a finely grained matrix of quartz and barite. Sample GASH ( Figure 10D) records the development of monazite at the expense of bastnaesite. Sample GASA ( Figure 10E) represents a deeply-altered ore which consists of brown relics of bastnaesite in a green monazitic groundmass. Sample MURA ( Figure 10F) shows the ultimate stage of alteration where bastnaesite is completely absent. (F) MURA sample shows a monazite (Mnz) ore associated with a quartz vein. Mineral abbreviations according to [27,28]. The black bars at the bottom right corners correspond to 2 cm.

Mineralogical Assemblages
The descriptions of the mineralogical assemblages are based on the observations made by optical microscopy (OM) and Scanning Electron Microscopy (SEM). Thin section OM pictures ( Figure 11A,B,E) display an assemblage of bastnaesite, monazite, quartz, biotite, barite and rhabdophane (hydrated REE phosphate). The SEM pictures reveal the presence of additional minerals such as galena, goethite and cerianite (Ce and Th oxide; Figure 11C,D,F).  [27,28]. As a summary, the mineralogical observations above, which are consistent with and complete the works of [17,29] are presented in a paragenetic sequence ( Figure 12). Bastnaesite and monazite are the most abundant minerals in the studied samples. Bastnaesite forms white tabular fractured crystals. Bastnaesite and the gangue minerals (barite, biotite) are affected by a dense network of micro veinlets of monazite and quartz. Biotite forms large brown crystals with perfect cleavage ( Figure 11A). Galena is present as inclusions in quartz ( Figure 11C) and presents a sharp contact with bastnaesite. Galena is the most abundant sulfide, however, traces of pyrite and molybdenite have been observed. Goethite which commonly forms a pseudomorph after pyrite was detected as dark crystals within geodes ( Figure 11D,E). Acicular crystals of rhabdophane form around quartz and replace bastnaesite and monazite ( Figure 11E). Cerianite grains locally replace bastnaesite and/or monazite; a second type of monazite, enriched in lanthanum (La) grew during or after the cerianite rims ( Figure 11F).
The REE mineralization consists of three successive stages ( Figure 12): The first stage consists of the deposition of primary ore comprising bastnaesite, biotite, galena, barite and quartz (1st generation). The second stage begins with the formation of monazite and quartz (2nd generation). The third stage corresponds to the formation of cerianite, a second type of La-rich monazite and rhabdophane in an oxidative environment (possibly linked to lateritization).

Materials and Methods
U-Th-Pb geochronology of monazite and bastnaesite grains from the Gakara deposit was conducted by in-situ laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at Geosciences Rennes using an ESI NWR193UC Excimer laser coupled to a quadripole Agilent 7700x ICP-MS directly in context in thin sections.
For monazite, the signals of 204 (Pb + Hg), 206 Pb, 207 Pb, 208 Pb, 232 Th and 238 U masses are acquired. No common Pb correction was applied owing to the large isobaric interference with 204 Hg. The 235 U signal is calculated from 238 U on the basis of the ratio 238 U/ 235 U = 137.88. Single analyses consisted of 20 s of background integration followed by 60 s integration with the laser firing and then a 10 s delay to wash out the previous sample. For more information on the settings of the instruments, see [30] and Appendix A Table A1. Spots diameters of 20 µm associated with repetition rates of 2 Hz with a laser fluency of 7.7 J/cm 2 were used during this study. Data were corrected for U-Pb and Th-Pb fractionation and for the mass bias by standard bracketing with repeated measurements of the Moacir monazite standard [31]. Data reduction was carried out with the GLITTER ® (v. 4.4.2) package developed by the Macquarie Research Ltd. [32]. Repeated analyses of the Manangoutry monazite standard (557.7 ± 3.1 Ma (MSWD = 0.40; N = 10); ID-TIMS age: 555 ± 2 Ma; [33] treated as unknown were used to control the reproducibility and accuracy of the corrections.
Bastnaesite has been successfully dated by U-Pb ID-TIMS by [34] who obtained a concordant age of 118 ± 1 Ma on two separate aliquots (no measurement of the Pb/Th ratio) for the Karasug Complex (Tuva Republic). This bastnaesite has been then used as in in-house standard (K-9) in the studies of [35] (in situ LA-ICPMS U-Pb dating) and [36] (in-situ SIMS Th-Pb dating). Unfortunately, we did not have access to this in-house standard which is, to our knowledge, the only one available. Indeed, the standard protocol for in-situ U-Th-Pb dating requires the use of a matrix-matched standard as it well known that this technique is highly affected by matrix effects (i.e., laser sampling may result in highly variable amounts of material delivered to the plasma torch, a phenomenon known as "ablation yields"; see [37] for more details). Therefore, in this study, we decided to standardize the bastnaesite grains against two different non-matrix-matched standards: A first set of measurements was standardized to the monazite standard Moacir (another LREE-bearing mineral), while the second set was standardized to the calcite standard WC-1 (another carbonate mineral).
The idea behind this analytical strategy was the following. If indeed this double-standardization protocol returns two apparent ages that are not comparable within error, this means that at least one (if not both) of the standardization did not correct properly for matrix-dependent effects and that, therefore, none of the obtained apparent ages can be considered as accurate. On the other hand, if both standardizations yield the same apparent age within error, the chances that non-matrix effects were not accounted for properly are minimized. Of course, we acknowledge the fact that this does not mean that we cannot be one hundred percent sure that the measured apparent ages are accurate. When bastnaesite analyses were standardized with the Moacir Monazite standard, we followed the same protocol as for the monazite dating (see above). For the standardization using the WC-1 calcite standard, we followed the recommendations of Roberts et al. [38], i.e., a spot size of 100 µm with repetition rates of 10 Hz and a laser fluency of 7.7 J.cm −2 . Data reduction was carried out using the data reduction scheme VizualAge_UcomPbine, a set of Iolite procedures that work with Igor Pro [39]. More information can be found in Appendix A Table A1 and in [40]. Repeated analyses of the WC −1 calcite standard ( 207 Pb corrected age of 257.7 ± 4.8 Ma (Common Pb 207 Pb/ 206 Pb = 0.85 ± 0.04; MSWD = 2.2; N = 5); ID-IRMS age 254.4 ± 6.4 Ma) [38] treated as unknown were used to control the reproducibility and accuracy of the corrections.
All concordia diagrams were generated using Isoplot/Ex (v. 3nx, BGC Berkeley Geochronology Center, Berkeley, CA, USA) [41]. All errors given in Tables 1 and 2 are listed at 1-sigma, but where data are combined to calculate lower intercept or concordia ages, the end results are provided at 2-sigma.

Monazite Dating
Fourteen monazite grains have been analyzed by LA-ICP-MS (Table 1). They are characterized by rather homogeneous, although low, Pb, U and Th contents (1.8-3.5 ppm, 14-30 ppm and 17-48 ppm respectively) and very consistent Th/U ratios (1.0 to 2.1). In Figure 13, we report the data in a Tera-Wasserburg (i.e., 207 Pb/ 206 Pb versus 238 U/ 206 Pb) Concordia diagram ( Figure 13A) and in a 206 Pb/ 238 U versus 208 Pb/ 232 Th Concordia diagram ( Figure 13B,C). In Figure 13A, the data plot in sub-concordant to discordant positions attributed to variable common Pb contents. They however yield a well-defined lower intercept date of 589.4 ± 6.1 Ma (MSWD = 1.5). In the 206 Pb/238U versus 208 Pb/ 232 Th Concordia diagram ( Figure 13B), the data are concordant (5 analyses) to discordant, demonstrating again the presence of common Pb in some of the analyzed grains. The five concordant data define a Concordia age (as of [42]) of 588.7 ± 7.5 Ma (MSWD = 0.73, Figure 13C) while all the data yield a lower intercept date of 585.5 ± 7.4 Ma (MSWD = 1.4, Figure 13B). Therefore, all the chronometers (U/Pb and Th/Pb) return dates that are comparable within error at ca. 588 Ma, a date that we interpret as the age of the crystallization of these monazite grains.

Bastnaesite Dating
As expose earlier, bastnaesite grains have been dated following a double-standardization protocol. They have been first standardized with a monazite standard. The data are listed in Table 2. Twenty-six spots have been analyzed. Pb, U and Th contents are fairly homogeneous with values comparable with the ones found for the monazite grains (1.2-9.6 ppm, 9-30 ppm and 13-59 ppm respectively). Their Th/U ratios are also comparable with the ones found for monazite, between 0.9 and 2.4. Although the calculation to obtain these contents is empirical (see [43], these values are comparable with the U and Th contents measured directly in the Bastanesite by LA-ICP-MS (U < 26 ppm and Th < 63 ppm).
Plotted in a Tera-Wasserburg diagram ( Figure 14A), they plot in a concordant to discordant position, due to the presence of variable amount of common Pb. Five analyses are concordant and yield a Concordia age of 608 ± 10 Ma ( Figure 14B; MSWD = 1.7). All the 26 analyses define a lower intercept date of 602.6 ± 9.5 Ma ( Figure 14A; MSWD = 2.3). When plotted in a 206 Pb/ 238 U versus 208 Pb/ 232 Th Concordia diagram ( Figure 14C), data are slightly discordant to discordant and define a lower intercept date of 597 ± 9 Ma (MSWD = 1.4).
A second set of measurements has been done using a calcite standard (Table 3). Because this calcite standard is almost exempt of any thorium (see [40]), only the U/Pb ratios have been measured. Twenty-six analyses have been performed and plotted in a Tera-Wasserburg diagram ( Figure 14D), they define a lower intercept date of 579 ± 27 Ma (MSWD = 0.27). If all the data acquired on bastnaesite (i.e., standardized with monazite and calcite) are plotted, they define a lower intercept date of 602.1 ± 7.2 Ma ( Figure 14E; MSWD = 1.5). This demonstrates that using two different non-matrix matching standards, the dates obtained for the bastnaesite are consistent with each other. We can therefore be relatively confident that the dates obtained are not biased by this analytical protocol and that bastnaesite crystallized 602.1 ± 7.2 Ma ago.

Sequence of Events
From the paragenetic sequence established in Figure 12 and in agreement with several previous works [3,17,29,44], it is clear that mineralization occurred in successive stages. The primary mineralization consisted in the deposition of bastnaesite, which was subsequently altered into monazite. A late oxidative stage leading to the newly-formation of La-rich monazite, cerianite and rhapdophane is likely related to lateritization, and will not be considered further in this discussion.
The main question remaining after the work of [6], which gave an age of 586.8 ± 3.7 Ma for bastnaesite crystallization, was the time elapsed between bastnaesite deposition and its alteration into monazite. Our geochronological data help addressing this point. The U-Th-Pb age obtained on monazite (588.7 ± 7.5 Ma) is close to the U-Th-Pb age of the bastnaesite (602.1 ± 7.2 Ma). Given the uncertainties obtained on the geochronological ages, it is difficult to distinguish in time the bastnaesite event (602 ± 8 Ma) from the monazitic alteration (589 ± 8 Ma). Also, our data are compatible with the paragenesis sequence where monazite crystallization postdates the bastnaesite crystallization. The fact that the La-Ba age on bastnaesite of [6] is similar to our U-Th-Pb age on monazite is difficult to discuss in the present study because we do not know much about the closing temperature of the La-Ba system regarding the bastnaesite mineral nor the behavior of this system when late fluids are involved. In any case, the main conclusion regarding the sequence of events is that the alteration stage followed the primary mineralization during a continuum of hydrothermal activity.
To explain the evolution from carbonated conditions to phosphatic ones, two hypotheses can already be raised considering the brief time elapsed between the two events. First, in the carbonatitic realm, it is known that phosphates can be concentrated in the late stages of evolution [16]. The Gakara case could therefore be one of these examples. Second, the fractures along which the mineralizing fluids circulated are typical of the upper brittle crust. Per descensum circulation of surface waters might thus be linked to the monazitic alteration and the associated silicification (2nd generation of quartz). The fluid inclusions data mentioned in [3] and the work of [4] indicate that the first hypothesis is the most realistic. Specific analyses like fluid inclusion characterization on the primary bastnaesite grains and secondary quartz associated with monazite and a stable isotope study of both bastnaesite and quartz must be undertaken to clarify the origin of the fluids involved in the Gakara system.

Structural Controls and Associated Fluid Pressure
First, the observed texture of the ore veins brings several instructive information. Fibers/unidirectional textures are rare, and it is therefore difficult to interpret the ore veins in term of fracturing modes (e.g., extensional vs. hybrid shear dilatant fractures). Whereas no ductile shearing has been observed, many ore veins present hydrothermal breccia texture, a feature more compatible with cataclasis and chemical corrosion during hydrothermal flow ( Figures 10C and 11B).
Second, the orientations and dips of the REE-bearing veins are highly variable between the different occurrences/prospects (see stereograms in Figures 3 and 7) as well as within a single occurrence. Together with the lack of fracturation evidences, it is therefore not possible at this stage to propose a regional stress field compatible with the REE-bearing vein network. This high variability can be partly linked to the late (post REE-vein emplacement) tectonic events, well identified at Kiyenzi (Figures 5 and 6), that could be responsible for bloc rotations/tilting, highly probable along the Tanganyika rift eastern shoulder.
However, various points must be discussed as they yield new insights on the hydro-tectonic context: (i) numerous and common, not to say systematic, evidences for reactivation and reuse of older mechanical heterogeneities (e.g., old Kibaran shear zones within gneisses, pegmatites, albitized aplites, etc.) during the REE mineralizing event. These reactivations lead to a spatial distribution of the REE-bearing vein strikes and dips that is more heterogeneous than what could be expected if the veins developed within an isotropic basement under a high regional differential stress field; (ii) The anastomosed REE-bearing vein pattern of the Gasagwe mine with curved 3D shapes, splay-off and thickness variations (Figure 3) is not compatible with a fracturing controlled by a high differential stress. This geometry resembles more that of hypovolcanic magmatic or phreatic-magmatic dykes and sills intruding shallow sediments, that propagated by opening anisotropies such as the beddings. Thus, the lack of a clear regional tectonic control for the REE-bearing vein distribution might also be explained by the injection of hot REE-rich fluids under a very low regional differential stress. Further works and data are required to constrain the structural model. This will be possible in the near future thanks to the mining activity bringing new outcrops; (iii) Independently of the structural model, hydro fracturing along the REE-bearing vein wall rocks (Figure 4), flat REE-bearing veins with large opening just under microgranite sills ( Figure 9A), evidence local supra-hydrostatic fluid pressure within the mineralizing plumbing system.

Geodynamic Context of the Gakara REE Ore Deposition
Several considerations allow to propose that the Gakara REE deposition occurred in the external zones of the Pan-African mobile belts. First, the geochronological data presented here undoubtedly assign a Pan-African age to the Gakara deposition. If compared to the paroxysmic tectonic and metamorphic events described for the Pan-African mobile belts in East Africa (the Mozambic mobile belt for example; see [7,9,44,45]), the age obtained in this study at ca. 600 Ma corresponds to a pre-collisional event, possibly in an extensional context, associated with the break-up of Rodinia (the interested reader may refer to the Figure 9 of [10], which provides a summary of the tectonic-metamorphic-magmatic events that prevailed since the Mid-Proterozoic in the east African region). Second, the Pan-African overprint is weak in Burundi as most of the eastern part of the country is covered by upper Proterozoic sediments [22,23]. Broadly, the contemporaneity between sedimentation and the deposition of the Gakara mineralization is consistent with an extensional regime for the region. Third, our data indicate the involvement of fluids coming from depth, in a good agreement with the carbonatitic-derived hydrothermal system classically invoked for Gakara. Carbonatite magmas generally emplace in a stable (i.e., non-tectonic) context or in association with lithospheric stretching. In passive margins, local thermal disequilibrium between the thinned and cold sub-continental lithospheric mantle and the nearby hot convective asthenospheric mantle leads to a weak partial melting of the former, producing low partial melting-derived melts, such as the alkaline and associated carbonatitic magmas, and the related REE-rich fluids.

Conclusions
Field observations of the Gakara REE deposits show that mineralization occurred in the upper brittle crust, in relation with primary fluids coming from depth. Fluids were localized along fractures zones and heterogeneities in the host rocks (including the steeply-dipping foliation, the walls of vertical aplitic veins). Where such heterogeneities were absent, over-pressure features are recorded by the ore bodies. The fact that we document hydrothermal fluids coming from below is consistent with the classical hypothesis of a hidden carbonatitic body at depth. The involvement of surface-derived fluids, notably during the alteration of bastnaesite into monazite, remains an open question at this stage; a stable isotope and fluid inclusions study is necessary to address this possibility. The ages obtained on bastnaesite (602 ± 7 Ma) and monazite (589 ± 8 Ma) demonstrate that bastnaesite crystallization was rapidly followed by monazite alteration, reinforcing the idea that the diverse types of ore bodies observed in the field must be reconciled into a unique mineralizing event, likely linked to the extensional tectonic context that prevailed at that time in the external zones of the Pan-African system.    (1) ) and 10 Hz (Cal (2) ) Spot size 20 µm (Mnz (1) ) and 100 µm (Cal (2) ) Sampling mode/pattern Single spot Carrier gas 100% He, Ar make-up gas and N 2 (3 mL/mn) combined using in-house smoothing device Background collection 20 s Ablation duration 60 s (Mnz (1) ) and 40 s (Cal (2)  Glitter [32] for Mnz (1) and Iolite [47] with VizualAge_UcomPbine [38] for Cal (2)

Uncertainty level & propagation
Propagation is by quadratic addition according to [48]. Reproducibility and age uncertainty of reference material are propagated.