The Puy-les-Vignes W deposit, located in the northwestern French Massif Central (FMC), is a rare occurrence of a wolframite-mineralized hydrothermal breccia pipe hosted in high-grade metamorphic gneisses. We present an integrated study of this deposit aiming to characterize the ore-forming hydrothermal system in link with the Variscan late-orogenic evolution of the FMC. Based on a set of representative samples from the host rocks and mineralization, we describe a detailed paragenetic sequence and we provide the major and trace element geochemistry of the granitic rocks and W–Nb–Ta–Sn–Ti oxide minerals, in situ U/Pb and 40Ar/39Ar geochronology, and a fluid inclusion study of quartz and wolframite. We demonstrate that the formation of this W-mineralized breccia pipe results from a multistage development related to four major episodes during the late Carboniferous. The first episode corresponds to the emplacement of an unexposed peraluminous granite at ca. 324 Ma, which generated microgranite dykes exposed at the present-day surface. The second episode is the formation of the quartz-supported breccia pipe and wolframite mineralization at ca. 318 Ma at a paleodepth of 7 km. The mineralizing fluids have a H2O–NaCl–CO2–CH4–N2 composition, a moderate-salinity (< 9 wt.% NaCl eq) and were trapped at high-temperatures (> 400 °C) during lithostatic to hydrostatic pressure variations caused by hydrofracturing of the host rocks. Wolframite deposition is interpreted to result from a W-rich intermediate-density magmatic fluid that exsolved from an evolved leucogranite and interacted with volatile-rich metasedimentary country rocks and/or possibly mixed with low-salinity metamorphic fluids of deep origin. The third episode corresponds to magmatic-hydrothermal Nb–Ta mineralization overprinting the W-mineralized system interpreted to be related to the intrusion at ca. 311 Ma of a rare-metal granite, which is part of a regional peraluminous rare-metal magmatism during the 315–310 Ma period. Finally, the last episode corresponds to disseminated Bi ± Au–Ag mineralization emplaced at ca. 300 Ma, which shares similar mineralogical features with late Carboniferous orogenic gold deposits in the FMC. The Puy-les-Vignes W deposit records, therefore, a multistage and long-lived development that extends over a timespan of 25 million years in a regional setting dominated by protracted peraluminous magmatism and high-temperature and low-pressure metamorphism. Although the local environment of ore deposition is atypical, our results show that the mineral assemblages, alteration styles, and fluid characteristics of the Puy-les-Vignes breccia pipe are similar to those of other peri-granitic W deposits in the FMC.


Le gîte à W de Puy-les-Vignes, situé dans le nord-ouest du Massif Central Français (MCF), est une occurrence rare de conduit bréchique hydrothermal minéralisé en wolframite et encaissé dans des gneiss de haut-degré métamorphique. Nous présentons une étude intégrée de ce gîte dans l’objectif de caractériser le système hydrothermal minéralisé en lien avec l’évolution tardi-orogénique varisque du MCF. Sur la base d’un ensemble d’échantillons représentatifs des roches encaissantes et de la minéralisation, nous décrivons une séquence paragénétique détaillée et nous présentons la géochimie des éléments majeurs et traces des roches granitiques et des oxydes de W–Nb–Ta–Sn–Ti, la géochronologie U/Pb et 40Ar/39Ar in situ, et une étude des inclusions fluides dans le quartz et la wolframite. Nous démontrons que la formation du conduit bréchique minéralisé en W résulte d’un développement multiphasé liée à quatre épisodes majeurs au cours du Carbonifère supérieur. Le premier épisode correspond à la mise en place d’un granite peralumineux non-affleurant à environ 324 Ma, qui a produit des dykes de microgranites affleurant à la surface actuelle. Le second épisode est la formation du conduit bréchique quartzeux et de la minéralisation à wolframite à environ 318 Ma à une paléo-profondeur de 7 km. Les fluides minéralisateurs ont une composition à H2O–NaCl–CO2–CH4–N2, une salinité modérée (< 9 wt.% NaCl eq) et ont été piégés à hautes températures (> 400 °C) pendant des variations de pressions lithostatiques à hydrostatiques causées par la fracturation hydraulique des roches encaissantes. Le dépôt de la wolframite est interprété comme le résultat d’un fluide magmatique riche en W de densité intermédiaire issu d’un leucogranite évolué qui a interagit avec des roches métasédimentaires riches en volatiles et/ou s’est possiblement mélangé avec des fluides métamorphiques de faible salinité d’origine profonde. Le troisième épisode correspond à la formation d’une minéralisation magmatique-hydrothermale à Nb–Ta se superposant au système minéralisé à W et lié à une intrusion de granite à métaux rares mise en place à environ 311 Ma, qui fait partie d’un magmatisme peralumineux à métaux rares d’échelle régionale pendant la période 315–310 Ma. Enfin, le dernier épisode correspond à une minéralisation disséminée à Bi ± Au–Ag mise en place à environ 300 Ma qui partage des similarités minéralogiques avec les gîtes d’or orogéniques tardi-Carbonifère du MCF. Le gîte à W de Puy-les-Vignes enregistre donc un développement multiphasé et de longue durée s’étalant sur 25 millions d’années dans un contexte régional dominé par du magmatisme peralumineux prolongé et du métamorphisme de haute température et basse pression. Bien que l’environnement de formation du gîte soit atypique, nos résultats montrent que les assemblages minéraux, les styles d’altérations et les caractéristiques des fluides du conduit bréchique de Puy-les-Vignes sont similaires à ceux d’autres gîtes péri-granitiques à W du MCF.


Tungsten and tin mineralization worldwide is spatially associated with granitoid intrusions forming a large variety of magmatic-hydrothermal ore deposits including greisens, veins, stockworks, skarns, pegmatites, porphyries, and breccia pipes (Černý et al., 2005; Romer and Kroner, 2016). Peri-granitic vein systems represent an economically important class of W–Sn deposits that are mostly spatially associated with metaluminous to peraluminous granitoids (Wood and Samson, 2000; Černý et al., 2005). The European Variscan belt is a large metallogenic province for W–Sn and rare-metal (Nb, Ta, Li, Be) deposits that are mainly located in Cornwall, the Erzgebirge, and the Iberian Massif (Romer and Kroner, 2016). In France, most W–Sn deposits are known in the French Massif Central (FMC; Marignac and Cuney, 1999), the Armorican Massif (Chauris and Marcoux, 1994), and the French Pyrenees (Poitrenaud et al., 2019). These deposits include wolframite-cassiterite-bearing quartz veins, greisen cupolas, and scheelite skarns spatially associated with Carboniferous peraluminous biotite-cordierite or two-mica granites (Stussi, 1989; Marignac and Cuney, 1999; Bouchot et al., 2005). Mining production for W in France was about 26 kt WO3 until 1986 and remaining resources are estimated at about 83 kt WO3 (Audion and Labbé, 2012).

The Puy-les-Vignes W deposit, located in the northwestern FMC, represents the only known wolframite-mineralized hydrothermal breccia pipe in France. Another occurrence of wolframite-bearing breccia pipe is known at Borralha, Portugal (Gonçalves et al., 2017; Bobos et al., 2018). Tourmaline breccia pipes are commonly associated with porphyry Cu deposits (Sillitoe, 1985; Skewes et al., 2003; Anderson et al., 2009), reduced intrusion-related Au deposits (Baker and Andrew, 1991; Thompson et al., 1999; Chen et al., 2009), but more rarely with granite-related W–Sn deposits (Ren et al., 1995; Somarin and Ashley, 2004; Yang and Bodnar, 2004; Solomovich et al., 2012). Magmatic-hydrothermal breccia pipes generally form cylindrical subvertical bodies in the upper part of shallow intrusions. They are generally interpreted as resulting from hydraulic brecciation and collapse of the host rocks caused by ascending overpressured magmatic fluids (Sillitoe, 1985; Skewes et al., 2003; Yang and Bodnar, 2004; Demirel et al., 2009). Release of magmatic fluids from crystallizing intrusions under specific pressure-temperature conditions produces greater mechanical energy and volumetric expansion of the host rocks resulting in the formation of breccia pipes associated with silicification and tourmalinization (Burnham, 1985; Pollard et al., 1987; Halls, 1994).

In this work, we present an integrated study of the Puy-les-Vignes W deposit aiming to characterize the formation of this atypical breccia pipe in link with the Variscan late-orogenic evolution of the FMC. Based on a set of representative samples from the host rocks and mineralization, we present a detailed paragenetic sequence of the deposit. Building on this sequence, we used a multi-method approach with the following objectives: (i) characterizing the major and trace element geochemistry of granitic rocks and W–Nb–Ta–Sn–Ti oxides, (ii) dating magmatic and hydrothermal events by in situ U/Pb and 40Ar/39Ar geochronology, and (iii) determining the origin and the conditions of the mineralizing fluids based on a fluid inclusion study of quartz and wolframite. We demonstrate that the formation of the Puy-les-Vignes breccia pipe results from a multistage development between ca. 325 and ca. 300 Ma in a regional setting dominated by protracted peraluminous granitic magmatism and high-temperature and low-pressure metamorphism.

Geological setting of the Puy-les-Vignes W deposit

The Puy-les-Vignes W deposit (Lat. 45°49’34”N, Long. 1°31’51”E) is located in the Limousin area in the northwestern part of the FMC (Fig. 1A), which belongs to the inner zone of the European Variscan belt formed during the continental collision between Gondwana and Laurussia through the Upper Paleozoic (Faure et al., 2009a; Lardeaux et al., 2014). The geologic structure of the FMC consists of a stack of three major metamorphic nappes that are from top to bottom (Faure et al., 2009a and references therein): (i) the Upper Gneiss Unit (UGU) composed of ortho- and paragneisses derived from Cambrian to Early Ordovician protoliths (530–470 Ma; Alexandrov et al., 2001; Berger et al., 2010; Chelle-Michou et al., 2017; Lotout et al., 2017, 2018) having experienced medium-pressure and medium-temperature amphibolite- to granulite-facies metamorphism at 360–340 Ma (Chelle-Michou et al., 2017); (ii) the Lower Gneiss Unit (LGU) consisting of ortho- and paragneisses similar to those of the UGU that were inherited from late Ediacarian protoliths (620–540 Ma; Alexandre, 2007; Melleton et al., 2010; Chelle-Michou et al., 2017; Couzinié et al., 2017, 2019); and (iii) the Para-Autochtonous Unit (PAU) that consists of low-grade metasedimentary rocks (micaschists, metagreywackes, and quartzites) that were derived from Neoproterozoic to Early Cambrian protoliths (650–530 Ma; Melleton et al., 2010). These metamorphic units were intruded by numerous granitoids during the Carboniferous between 360 and 300 Ma (Faure et al., 2009a; Lardeaux et al., 2014; Couzinié et al., 2017; Laurent et al., 2017; Moyen et al., 2017; Villaros et al., 2018).

The Puy-les-Vignes breccia pipe is hosted in migmatitic biotite-sillimanite paragneisses of the Saint-Léonard-de-Noblat gneiss unit, which is structurally correlated with the LGU and is located on the southern flank of the Thaurion anticline (Fig. 1B). The deposit is situated 20 km south of the Saint-Sylvestre leucogranite complex dated at 324 ± 4 Ma (Holliger et al., 1986; Cuney et al., 1990) and 15 km west of the Millevaches leucogranite complex emplaced between 335 Ma and 325 Ma (Rolin et al., 2014). The breccia pipe is close to two peraluminous granites (Fig. 2A): the Auriat two-mica granite in the northeast dated at 324 ± 1 Ma (Gebauer et al., 1981), and the Aureil biotite-cordierite granite in the southwest dated at 346 ± 14 Ma (Duthou, 1978). The breccia pipe has dimensions of 80 × 340 m at surface and is cut in two parts (west and east) by a N30°E-trending sinistral strike-slip fault (“Faille Limite”) with a 120-m horizontal displacement (Figs. 2B and 2C; Weppe, 1951, 1958). Only the eastern part of the breccia pipe was mined down to a depth of 300 m whereas the western part was barely explored. The vertical extent of the deposit is unknown, but an unexposed granite may exist at depth as described for other mineralized breccia pipes elsewhere (Sillitoe, 1985; Ren et al., 1995; Somarin and Ashley, 2004; Yang and Bodnar, 2004; Solomovich et al., 2012). Regional geophysical data reveal a low gravity anomaly with a NE-SW elongation around Puy-les-Vignes, suggesting a possible extension of the Auriat granite beneath the breccia pipe (Melleton et al., 2017). Microgranite and lamprophyre dykes, of 1–4 m thickness and NE-trending, are located at the intersection with the breccia pipe and are fragmented within the breccia (Fig. 2B; Weppe, 1951, 1958). Field relationships indicate that the microgranite postdate the lamprophyre. At depth, another NE-trending lamprophyre dyke crosscuts the breccia pipe possibly implying the existence of several generations of lamprophyre (Weppe, 1958). Similar lamprophyre dykes crosscutting the Saint-Sylvestre and Millevaches leucogranitic complexes yielded Rb/Sr whole rock ages of 285 ± 10 Ma (Leroy and Sonet, 1976) and 290 ± 5 Ma (Chalier et al., 1994), respectively.

From a regional perspective, the Puy-les-Vignes deposit is centered on an extended W–As geochemical stream-sediments anomaly (> 60 ppm W; Bertrand et al., 2015). Peri-granitic W ± Sn deposits occur regionally such as Mandelesse (Guyonnaud et al., 1977), Saint-Goussaud (Alikouss, 1993), and Vaulry (Vallance et al., 2001). Several W occurrences are known within a distance of < 2 km around the Puy-les-Vignes deposit, including quartz-wolframite veins and stockworks at the localities Lifarnet, Le Moulard, Les Clauds, Etivaud, Chassagnat and Beynat, as well as a second possible wolframite-bearing breccia pipe at Les Caillaudoux (Weppe, 1951, 1958; Fig. 2A). Additional quartz-wolframite veins have been documented in an area between Cheissoux and Villetelle, as well as a greisen-like cupola at Moissannes, respectively located at 15 km east and 10 km north of Puy-les-Vignes (Dudek, 1978). The Puy-les-Vignes deposit was mined underground between the mid 19th and 20th centuries (1855–1957) with a cumulated production of 3733 t WO3 at an average grade of 0.5% WO3 (Mouthier, 2005) and was the second-ranked tungsten mine in France during the 20th century. Proved reserves are estimated at about 1158 t WO3 based on mining archives (Melleton et al., 2017), but 3D numerical geologic modelling re-estimated reserves of 2500 t WO3 for the eastern part of the breccia pipe by considering a vertical extent of 235 m and an average grade of 0.2% WO3 (J.J Royer, personal communication). Peñarroya, after a diamond drilling campaign in the 1970’s on the unexploited western part of the Puy-les-Vignes breccia pipe, reported the discovery of a 4-m length intercept at 3% WO3 (Melleton et al., 2017). The potential for discovering new W resources is significant at Puy-les-Vignes considering the number of wolframite occurrences in the district, that only half of the breccia pipe was mined, that the vertical extent of the deposit is unknown, and that the whole breccia body represents a potential low-grade ore.

According to Weppe (1951, 1958), the formation of the Puy-les-Vignes breccia pipe would result from three successive stages: (i) A first stage of hydraulic fracturing and mass collapse produced a matrix-supported breccia composed of polygenic and heterometric clasts of the country rocks (mainly gneisses) cemented by quartz constituting about 10% of the total volume of the breccia. Metric-size clasts have undergone syn-collapse rotation as suggested by the progressive tipping of gneissic blocks within the breccia where fragments adjacent to the pipe margins are subvertical and those towards the center are subhorizontal (Fig. 2D); (ii) A second collapse stage led to the formation of subvertical quartz-filled decollements (“gaine” and “croiseurs”), delimiting the borders of the breccia pipe with the country rocks, and quartz-filled planar-curved veins and dome structures (“plateures”) dipping at low-angle (< 30°) within the breccia. The contact between the breccia and the enclosing gneisses is sharp and consists of a 0.5-to-3-m thick quartz vein delimiting the southern and northern boundaries of the pipe; (iii) A third stage yielded the formation of a NE-trending network of subparallel quartz-wolframite-sulfide veins (“filons”) dipping at high-angle (> 50°) and crosscutting the entire breccia pipe and the host gneisses. These veins are about 400 m long with variable thickness of 0.2–1 m and contain most of the W mineralization with grades ranging from 0.2 to 2% WO3 (Mouthier, 2005). The quartz-filled decollements at the breccia-country rock boundary and the quartz cementing the breccia also contain wolframite but at lower grades (Fig. 3D). The W mineralization is dated at 317.7 ± 0.7 Ma based on U/Pb dating of wolframite from a mineralized vein (Harlaux et al., 2018a).

Materials and analytical methods


The Puy-les-Vignes mine is closed since 1957 and the underground workings are no more accessible. Previous work from Alikouss (1993) was based on a few samples taken from the historical collection of M. Weppe coming from the underground workings and from samples collected in mining dumps, which are less accessible today due to revegetalization. The present work relies on access to private collections of old mine workers (P. Medda, G. Pradeau) and to museum collections (Musée Minéralogique de l’École des Mines d’Alès, France, Collection P. Fitte) corresponding to samples coming directly from the underground mine when it was still operating. Information on the precise depth location and underground adit was available for some samples. Mining archives indicate that only ores from the Puy-les-Vignes mine were processed on site (Mouthier, 2005), thus excluding any risks of mixing with ores coming from other mines. Complementary samples for this study come from surface sampling of stripped terrains and trenches during recent site reworking and from a small outcrop of the breccia (Fig. 3). The latter is located at the margin of the eastern part of the breccia pipe close to a formerly mined quartz-wolframite vein (“filon 1”) in proximity to the old mining headframe (“Puits Girod”). An old underground adit of about 150 m in length cutting through the Saint-Léonard-de-Noblat gneisses and intersecting quartz-wolframite-sulfide-tourmaline veins was also studied at Lifarnet (< 2 km north of Puy-les-Vignes). This diversified sampling offers a set of representative samples for the host rocks and mineralization from the Puy-les-Vignes deposit (Fig. 4).

Petrography and geochemistry

Mineralogical analyses were performed at GeoRessources, Université de Lorraine (Nancy, France), using an Olympus BX51 optical microscope and a JEOL J7600F scanning electron microscope (SEM) equipped with a SDD-type energy dispersive X-ray spectrometer (EDS) coupled to an Oxford Wave wavelength dispersive X-ray spectrometer (WDS). Backscattered electron (BSE) images were acquired on carbon-coated polished thin sections with an acceleration voltage of 15 kV. Chemical compositions of Nb–Ta–W–Ti–Sn oxides were measured at GeoRessources using a CAMECA SX100 electron microprobe analyzer (EMPA) operated with a 20 kV accelerating voltage, a 20 nA beam current, and a 1 μm beam diameter. The following standards and elements were analyzed using WDS spectrometers: hematite (Fe Kα), MnTiO3 (Mn Kα, Ti Kα), olivine (Mg Kα), corundum (Al Kα), albite (Si Kα, Na Kα), Sc metal (Sc Kα), chromite (Cr Kα), LiNbO3 (Nb Lα), LiTaO3 (Ta Lα), scheelite (W Lα), cassiterite (Sn Lα), uraninite (U Mα), galena (Pb Mα). Counting times were 10 s on element peaks and 5 s on background. Minor and trace elements in rutile were measured by laser ablation – inductively coupled plasma – mass spectrometry (LA-ICP-MS) at GeoRessources using an Agilent 7500c quadrupole ICP-MS coupled with a 193 nm GeoLas ArF Excimer laser (MicroLas, Göttingen, Germany). Laser ablation was performed on polished thin sections with a constant 5 Hz pulse frequency and a fluence of 7 J cm−2 using variable laser spot diameters between 32 and 60 μm. Helium was used as carrier gas to transport the laser-generated particles from the ablation cell to the ICP-MS and argon was added as an auxiliary gas via a flow adapter before the ICP torch. Typical flow rates of 0.5 L min−1 for He and 0.9 L min−1 for Ar were used. The certified reference material NIST SRM 610 (Jochum et al., 2011) was used as an external standard for calibration of all analyses and was analyzed twice at the beginning and at the end of each set of samples, following a bracketing standardization procedure. The reference material NIST SRM 612 was used as a control standard for the standardization. Calibration of the ICP-MS was optimized for highest sensitivity on an intermediate m/Q range, while maintaining Th/U ∼ 1 and ThO+/Th+ < 0.5%, as determined on NIST SRM 610. The following isotopes were measured with a dwell time of 20 ms for each: 29Si, 45Sc, 47Ti, 51V, 53Cr, 57Fe, 60Ni, 63Cu, 66Zn, 90Zr, 93Nb, 95Mo, 118Sn, 181Ta, and 182W. The total cycle time was of 332 ms. Data reduction and absolute quantification of signals were performed using the software StalQuant (Fricker, 2012) developed at ETH Zürich (Switzerland) and the Ti content determined by EMPA. Limits of detection (LOD) were calculated using the 3σ criterion of Longerich et al. (1996). Whole-rock geochemistry of selected samples was determined at the SARM-CNRS (Vandoeuvre-lès-Nancy, France) following the procedure described by Carignan et al. (2001).

U/Pb geochronology

U/Pb dating of rutile was performed at the GeOHeLiS analytical platform (Géosciences Rennes/OSUR, University of Rennes, France) using an ESI NWR193UC Excimer laser system coupled to an Agilent 7700x quadrupole ICP-MS. The rutile grains were previously investigated by SEM and EMPA in order to select chemically homogeneous zones that lack microinclusions and alteration features. Laser ablation was performed on polished thin sections at repetition rate of 5 Hz, a fluence of 8.6 J cm−2, and using spot diameters of 40 or 50 μm depending on grain size. The ablated material was carried by He and then mixed with N2 and Ar (Paquette et al., 2014) before injection into the ICP torch equipped with a dual pumping system to enhance sensitivity. Typical gas flows of 0.75 L min−1 for He, 3 mL min−1 for N2 and 0.85 L min−1 for Ar were used. Tuning of the instrument and mass calibration were performed before the analytical session using the NIST SRM 612 reference material (Jochum et al., 2011) by monitoring the 238U signal and minimizing the ThO+/Th+ ratio < 0.5%. The signals of 204Pb, 206Pb, 207Pb, 208Pb, 232Th, and 238U were measured with a dwell time of 10 ms (30 ms for 207Pb). For each analysis, 20 s of background was integrated, followed by 60 s of laser-generated signal acquisition. Each analysis was separated by a 15 s delay of wash-out. Data were corrected for U/Pb fractionation and for the mass bias by standard bracketing with repeated measurements of the R10 rutile (1090 Ma; Luvizotto et al., 2009) reference standard. The R19 rutile reference material (489.5 ± 0.9 Ma; Zack et al., 2011) was analyzed as a secondary standard to control the reproducibility and accuracy of the corrections and yielded an age of 492.9 ± 4.4 Ma (MSWD = 0.89, n = 13). The U/Pb isotopic analyses of the rutile standards are reported in Supplementary Table 1. Data reduction and age calculations were carried out with the GLITTER software package (Van Achterbergh et al., 2001). Analyses yielding non-radiogenic isotopic ratios were rejected. Data were plotted using ISOPLOT v.3.75 (Ludwig, 2008). Error propagation is by quadratic addition according to Horstwood et al. (2016). For more information on the analytical procedure, see Boutin et al. (2016) and Supplementary Table 1.

40Ar/39Ar geochronology

40Ar/39Ar analytical work was performed at the University of Manitoba (Canada) using a multi-collector Thermo Fisher Scientific ARGUS VI mass spectrometer, linked to a stainless steel Thermo Fisher Scientific extraction/purification line, Photon Machines (55 W) Fusions 10.6 CO2 laser, and Photon Machines (Analyte Excite) 193 nm laser. Argon isotopes (from mass 40 to 37) were measured using Faraday detectors with low noise 1012 Ω resistors and mass 36 was measured using a compact discrete dynode (CDD) detector. All specimens (samples and standards) were irradiated for 17 h in the Cadmium-lined, in-core CLICIT facility of the TRIGA reactor at the Oregon State University, USA. Standards of Fish Canyon sanidine (Kuiper et al., 2008) and GA1550 biotite (Spell and McDougall, 2003) were placed in a Cu sample tray, with a KBr cover slip, in a stainless-steel chamber with a differentially pumped ZnS viewport attached to a Thermo Fisher Scientific extraction/purification line and baked with an infrared lamp for 24 h. Unknowns consist of discs of 5 mm in diameter and ∼ 150 μm thick that were cut from thick polished section investigated by optical microscopy and SEM prior to 40Ar/39Ar analyses. The discs were mounted using a ceramic adhesive (PELCO) on a quartz slide placed in a stainless-steel chamber with a sapphire viewport attached to the same stainless-steel high vacuum extraction system as the CO2 laser, and baked with an infrared lamp for 48 h. For this study, a raster size of about 100 μm × 100 μm was used and ablation pits were excavated to an estimated depth of 50 μm. Five argon isotopes were measured simultaneously over a period of 6 minutes. Measured isotope abundances were corrected for extraction-line blanks, which were determined before every sample analysis. Line blanks in both the Excimer and CO2 system averaged ∼ 3 fA for mass 40 and ∼ 0.013 fA for mass 36. Detector intercalibration (IC) between the different faraday cups was monitored (in Qtegra) every four days by peak hopping 40Ar. Calculated values are ICH1: 1.0000, ICAX: 1.0745, ICL1: 1.0637, and ICL2: 1.0534, with an error of ∼0.2%. The intercalibration factor between H1 and the CDD was measured using 70 air aliquots interspersed with the unknowns resulting in ICCDD: 1.0081 ± 0.0002 per amu. A value of 295.5 was used for the atmospheric 40Ar/36Ar ratio (Steiger and Jäger, 1977) for the purposes of routine measurement of mass spectrometer discrimination using air aliquots, and correction for atmospheric argon in the 40Ar/39Ar age calculation. Corrections are made for neutron-induced 40Ar from potassium, 39Ar and 36Ar from calcium, and 36Ar from chlorine (Renne et al., 1998; Renne and Norman, 2001). Data collection, reduction, error propagation and age calculation were performed using the MassSpec software (v.8.091; Deino, 2013). The decay constants used were those recommended by Steiger and Jäger (1977). Data were plotted using DensityPlotter (Vermeesch, 2012), and uncertainties in tables and text are quoted at the 1σ and 2σ level, respectively.

Fluid inclusion analysis

Fluid inclusions were studied using double-polished 150–200 μm thick sections of quartz-wolframite vein samples. Petrographic and microthermometric analyses of quartz-hosted fluid inclusions were performed at GeoRessources on a THMS 600 Linkam heating-cooling stage mounted on an Olympus BX-51 microscope. The stage was calibrated at −56.6 °C, 0.0 °C, and 374.1 °C using pure H2O and H2O–CO2 synthetic fluid inclusion standards. Petrography and microthermometry of wolframite-hosted fluid inclusions were carried out at the University of Geneva (Switzerland) using a Linkam FTIR 600 stage mounted on an Olympus BH-51 microscope equipped with an Olympus XM-10 near-infrared (NIR) camera (Ortelli et al., 2018). Calibration of the heating-cooling stage was performed regularly using SynFlinc standards at −56.6 °C, 0.0 °C, and 374.1 °C (Sterner and Bodnar, 1984). The visible light absorbed by opaque minerals is partially converted into heat in proportion to the opacity and this could potentially preclude accurate microthermometric measurements (Moritz, 2006). To avoid this problem, a voltmeter was connected directly to the lamp of the microscope to control the power of the incident light following the procedure described by Ortelli et al. (2018). Sample heating is not high enough to affect microthermometric measurements if a 90% closed diaphragm, a condenser at 0.3, and a power of < 4 V depending on the transparency of the different zones of the crystals are used (Casanova et al., 2018). The salinities, densities, and bulk compositions of CO2-bearing fluid inclusions were calculated based on the clathrate melting temperature (Tm clat) and homogenization temperature of the CO2 phase (Th CO2) using the Q2 program developed by Bakker (1997) and the equations of state from Thiéry et al. (1994) and Duan et al. (1996). Isochores were calculated using the ISOC program (Bakker, 2003) based on the equations of state from Bakker (1999). For H2O–NaCl fluid inclusions, the salinities, densities, and isochores were calculated based on the ice-melting temperature (Tm ice) and total homogenization temperature (Th) using the HokieFlincs_H2O–NaCl program (Steele-MacInnis et al., 2012). Raman spectroscopy analyses of fluid inclusions were carried out at GeoRessources using a Horiba Jobin-Yvon LabRAM HR spectrometer equipped with a 1800 g mm−1 grating and an Edge filter. The excitation beam is provided by a Stabilite 2017 Ar+ laser (Spectra Physics, Newport Corporation) at 514.532 nm and a power of 400 mW focused on the sample using x50 and x100 objectives mounted on an Olympus BX-40 optical microscope. The signal-to-noise ratio was optimized, ideally lower than 1%, by adjusting the acquisition time and accumulation number. Gas compositions of fluid inclusions were measured at room temperature using a 500 μm confocal hole aperture and a 100 μm slit aperture. The relative proportions of the gas species (mol%) were calculated using the peak area, acquisition time, and the specific Raman scattering cross-sections (Burke, 2001).

Paragenetic sequence of the Puy-les-Vignes deposit

Petrographic observations from the outcrop- to the micrometer-scale, completing previous descriptions (Weppe, 1951, 1958; Alikouss, 1993; Harlaux et al., 2015a; Harlaux, 2016), allow proposing a detailed paragenetic sequence of the Puy-les-Vignes deposit divided into six main stages (Fig. 5).

Stage I: Early hydrothermal alteration and brecciation

Greisenization of the country rocks (mainly gneisses) is the earliest hydrothermal alteration episode identified at Puy-les-Vignes. It corresponds to dissolution of primary feldspar and biotite in the Saint-Léonard-de-Noblat paragneisses and precipitation of secondary quartz, muscovite, and tourmaline (“Tur 1” of Harlaux et al., 2019; Fig. 6A). The foliation is partly preserved within the gneissic clasts (Figs. 3B and 6A) and relics of biotite and feldspar remain in unaltered parts of the clasts, suggesting that greisenization was diffuse and limited to structural discontinuities. Greisenization (stage Ia) was followed by hydraulic brecciation and collapse of the host rocks resulting in a clast-supported tourmaline crackle breccia (stage Ib). The clasts of gneisses and granites are angular to subangular in shape, centimetric to plurimetric in size, and lack granulometric classification (Fig. 3A). They are surrounded by an irregular rim of tourmalinite (tourmaline ± quartz) of variable thickness, ranging from a few millimeters to several centimeters (Figs. 3B, 3C, 4A and 4B), and up to complete tourmalinization of the clasts. Hydrothermal alteration of the clasts results from two successive steps: (i) chloritization of residual biotite associated with formation of Fe–W-rich rutile (Fig. 7A), monazite, xenotime, and zircon; (ii) tourmalinization of the clasts rims forming dense aggregates of prismatic tourmaline (“Tur 2” of Harlaux et al., 2019) cemented by quartz (Fig. 6B).

Stage II: W–(Sn)–As–Nb–Ta mineralization

Hydraulic brecciation and fluid-assisted reopening of the tourmaline crackle breccia produced the quartz-supported breccia pipe and wolframite-bearing quartz veins. The contact between the tourmaline-bordered clasts and the quartz cement is sharp and crack-seal textures are common (Figs. 4A and 4B). Syn-collapse rotation of the clasts occurred prior or concomitantly to quartz deposition as indicated by variable orientation of the foliation within gneissic clasts. Quartz crystals of up to 20 cm-length are locally found in geodic cavities within the breccia matrix. At the microscopic scale, quartz is composed of millimetric polycrystalline aggregates showing evidence of ductile deformation (undulose extinction, lamellar deformation) and dynamic recrystallization (bulging and subgrain rotation; Fig. 6C). Besides quartz, wolframite and arsenopyrite are the main minerals in the mineralized veins (Figs. 4C and 4D). Wolframite is euhedral ranging from millimetric to pluri-centimetric in size, exceptionally up to decimetric, and grew directly onto the clast rims. Most wolframite is Fe-dominated (Fe/(Fe + Mn) = 0.80–0.86; Harlaux et al., 2018b) but is locally replaced by a variety of wolframite having almost pure ferberitic composition (ferberite-1; Fe/(Fe + Mn) > 0.95) along microcracks and crystal borders (Figs. 6D and 7B). Arsenopyrite is the main sulfide in the quartz veins and is associated with minor pyrrhotite and crosscuts wolframite (Fig. 6E). Most arsenopyrite is characterized by a weight ratio of As/S = 2.5 but an early variety of arsenopyrite containing about 4.5 wt.% Co and 1 wt.% Ni with a weight ratio of As/S = 3 was identified by EDS analyses (Fig. 7C). Minor cassiterite is intergrown with arsenopyrite (Fig. 7D) and rare Cr-rich phengites (“mariposite”) are also found in association with arsenopyrite. Accessory minerals of tourmaline, muscovite, and Nb–Ta–W oxides are also present in the quartz veins. Prismatic acicular tourmaline (“Tur 3” of Harlaux et al., 2019) occurs perpendicularly on clast rims and as disseminations in quartz and is also locally intergrown with wolframite. Muscovite consists of millimetric flakes disseminated in quartz or overgrown on wolframite, and occasionally shows kink-band deformation (Fig. 6C). Accessory Nb–Ta–W oxide minerals associated with the stage II include (i) Nb–Ta-rich rutile disseminated in quartz (Fig. 7E), (ii) Fe–Nb-rich rutile disseminated within ferberite-1 (Fig. 7F), (iii) Ta–Nb-rich ferberite (1.6 wt.% Ta, 0.4 wt.% Nb, and 0.8 wt.% P; EDS analyses) in dissolution microvugs within wolframite (Fig. 7G), (iv) a variety of W-rich ixiolite (hereafter named wolframoixiolite) in quartz-filled veinlets cutting wolframite (Fig. 7H) and as disseminations overgrown by ferberite-1 (Fig. 7I), and (v) Cr–Nb–W-rich rutile disseminated within arsenopyrite (Fig. 7J) together with inclusions of Fe-skutterudite ([Fe,Co,Ni]As3), wolframite, xenotime, and monazite. Scheelite is late relative to the Nb–Ta–W oxides and occurs as centimetric bipyramidal euhedral grains having orange-brownish-colors (Fig. 4E). Scheelite partly replaces wolframite and ferberite-1 (Fig. 6F), crosscuts arsenopyrite along veinlets, and shows undulose extinction similar to quartz. A third generation of wolframite of ferberitic composition (ferberite-2; Fe/(Fe + Mn) > 0.95) occurs as pseudomorphs after ferberite-1 and scheelite and forms aggregates of randomly oriented fibrous crystals (Figs. 6G and 6H). Locally, ferberite-2 forms octahedral crystals of pluri-centimetric size in quartz geodic cavities (Fig. 4F), corresponding to the “reinite” described by Weppe (1951, 1958).

Stage III: Fe–Cu–Zn ± Mo mineralization

Base metal sulfide assemblages, including pyrite, chalcopyrite, and sphalerite overprint the mineral paragenesis of stages I and II. Pyrite and chalcopyrite are volumetrically the most important minerals of this stage. They crosscut wolframite and arsenopyrite (Figs. 6I6K) and are coeval with the deposition of quartz and chlorite (Fig. 7K). Sphalerite is early relative to pyrite and chalcopyrite (Fig. 6L) and contains up to 2 wt.% Cd, as determined by EDS analyses. Minor tennantite (identified by EDS) is associated with chalcopyrite (Figs. 6I and 6J). Molybdenite is thought to be coeval with chalcopyrite but its paragenetic position remains uncertain. Stannite is late and occurs locally as exsolution trails within chalcopyrite (Fig. 7L), as revealed by EDS analyses. Deposition of sulfides is associated with recrystallization of quartz along the crystal borders of wolframite and arsenopyrite. The sulfide assemblage and associated quartz are not deformed, thus suggesting that stage III postdates the main ductile deformation episode.

Stage IV: Nb–Y–HREE mineralization

This stage is recorded by a millimeter-scale hydrothermal crackle breccia composed of clasts of tourmaline (“Tur 4” of Harlaux et al., 2019) and vermicular chlorite cemented by adularia and overprinting a deformed quartz-wolframite-scheelite-sulfide vein (Fig. 6M). Adularia is not deformed indicating that the brecciation episode occurred after the main ductile deformation that affected the deposit. The crackle breccia contains a series of accessory minerals disseminated within adularia, including monazite, xenotime, zircon, Nb–Fe–W-rich rutile, and Nb–Ti–Y–HREE–W–U oxides referred to as NTox (Fig. 6N). The latter minerals were studied in detail by Harlaux et al. (2015b) and consist of subeuhedral grains (10–400 μm in size) having complex internal textures and elevated concentrations of Nb, Ti, Y, HREE, and W. This mineral paragenesis is characterized by a P, Y, HREE, Nb>Ta, Ti, Zr, and U geochemical signature, which has been interpreted by Harlaux et al. (2015b) to result from rare-metal-bearing hydrothermal fluids derived from an unknown peralkaline magmatic body at depth.

Stage V: Bi ± Au–Ag mineralization

A volumetrically minor assemblage of native bismuth, bismuthinite (Bi2S3), Bi–Ag–Se–Te sulfosalts, and electrum (ca. 78 wt.% Au and 22 wt.% Ag) fills microcavities and microcracks overprinting quartz, wolframite and arsenopyrite from stage II, base metal sulfides from stage III, and adularia from stage IV (Figs. 7I7M). Accessory Bi–Ag–Se–Te sulfosalt minerals include hedleyite (Bi7Te3) and sztrokayite (Bi3TeS2). Quartz, siderite, arsenopyrite, and colloform As-rich pyrite (1–10 wt.% As) precede the deposition of the Bi minerals (Figs. 6O and 7N7P). Russellite (Bi2WO6) locally occurs as pseudomorph after ferberite at the contact with native bismuth.

Stage VI: Late hydrothermal and supergene alterations

Late hydrothermal fluid circulations are responsible for alteration of arsenopyrite to scorodite (FeAsO4.2H2O) and other arsenates such as pharmacosiderite, carminite, and beudantite. Scorodite is usually greenish in color, has a skeletal porous texture, and occurs as pseudomorph after arsenopyrite or as veinlets crosscutting wolframite and chalcopyrite (Fig. 6P). Minor illite and rooseveltite (BiAsO4) are associated with scorodite, indicating the remobilization of bismuth during alteration of arsenopyrite. Supergene weathering of the orebody resulted in the partial destabilization of wolframite and residual sulfides in secondary minerals such as tungstite (WO3.H2O), colloform goethite, and Fe–oxides–hydroxides.

Petrography and geochemistry of granitic rocks

Microgranite dykes

Microgranite dykes crosscut the host paragneisses and are fragmented within the breccia and cut by tourmaline veins, indicating an emplacement prior to the early hydrothermal alteration and brecciation stages. The least altered sample of microgranite (sample PV-92-7), collected at surface, is a fine-grained rock composed of quartz, K-feldspar, plagioclase and muscovite (Figs. 4G and 8A). Quartz shows undulose extinction while feldspars are partly sericitized. The microgranite composition is strongly peraluminous (A/CNK = 1.54) and plots along the fractionation trend of evolved leucogranites, rare-metal granites (RMG), and rare-metal pegmatites (RMP) from the northwestern FMC (Fig. 9; Tab. 1). Its trace element composition overlaps those of muscovite-bearing peraluminous granitoids from the northwest FMC and is characterized by low contents of Sr (95 ppm), Th (0.5 ppm), Zr (22 ppm), Hf (0.9 ppm), and Pb (5.8 ppm), and high contents of Rb (156 ppm), W (9.7 ppm) and Sn (41 ppm), compared to the upper continental crust (Fig. 10A). The REE contents are low (11 ppm) and show a relatively flat spectrum (La/Yb = 1.38) with a positive Eu anomaly (Eu/Eu* = 1.66), which differs from the regional leucogranites (Fig. 10C). A strongly altered sample of microgranite (sample PV-92-5), also collected at surface, shows higher contents of Fe2O3 (2.9 vs. 0.5 wt.%) and K2O (4.3 vs. 2.5 wt.%) and lower contents of SiO2 (75.2 vs. 78 wt.%) and Na2O (0.2 vs. 3.6 wt.%) compared to the unaltered precursor. The trace element composition is similar to the least altered microgranite (Figs. 10A10C), except for W (30 ppm), Sn (98 ppm), Bi (34 ppm), As (3540 ppm), and Cu (246 ppm) reflecting the effect of hydrothermal alteration.

Granitic clasts

The granitic clasts within the breccia are strongly greisenized and tourmalinized, indicating that the granite emplacement preceded the sequence of hydrothermal events. Two samples of strongly altered granitic clasts were collected at surface (samples PLV-02-10 and PLV-13-02). These rocks are dominantly composed of secondary quartz and muscovite that replaced the primary feldspars. The altered granitic clasts have strongly peraluminous compositions (A/CNK = 3.00–3.02) that plot along greisenization trend in the geochemical diagrams together with the strongly altered microgranite sample (Fig. 9; Tab. 1). Their trace element compositions partly overlap those of muscovite-bearing peraluminous granitoids from the northwestern FMC (Figs. 10A10C). The REE spectra of the altered granites show a fractionated pattern (La/Yb = 1.74–6.36) with negative Eu anomalies (Eu/Eu* = 0.44–0.68). Expectedly, these altered granitic rocks have elevated concentrations of W (40–97 ppm), Sn (56–89 ppm), As (2235–2275 ppm), Bi (26–28 ppm), and Cu (71–301 ppm) as a result of hydrothermal alteration by mineralizing fluids.

Rare-metal granite

A unique sample of RMG coming from the Puy-les-Vignes mine has been studied (sample PLV-02-02). The location of this granite as well as the relations with the breccia are unknown, but it is certain that this sample was collected in the old underground workings. The RMG is a fine-grained leucocratic rock showing a magmatic layering, which consists of a repetitive sequence of quartz-albite-muscovite band (“B1” on Figs. 4H and 8B) progressively evolving to albite-quartz band (“B2” on Figs. 4H and 8C). The transition between the two zones is marked by a progressively decreasing grain size from 500–1000 μm on average in the quartz-albite-muscovite band to 100–200 μm in the albite-quartz band. Quartz consists of xenomorphic globular phenocrysts containing poikilitic inclusions of albite laths in a typical “snowball” helicitic texture (Fig. 8D) while the granitic groundmass is characteristic of fractionated albitic cumulates. There is no evidence of post-magmatic deformation and hydrothermal alteration, thus suggesting that the RMG emplaced after the main episode of ductile deformation postdating the paragenetic stage II. Muscovites contain elevated Si/Al ratios (1.12–1.97, avg = 1.32) and high contents of F (2.3–8.5 wt.%, avg = 4.2 wt.%), as determined by EDS analyses. Accessory minerals include apatite, cassiterite, and columbite-tantalite that are disseminated in the albite-quartz bands. No magmatic zircons and monazite were found. The lower part of the sample shows unidirectional solidification textures (UST) composed of perpendicularly oriented K-feldspar phenocrysts relative to the magmatic layering. The albite-quartz band (referred as PLV-02-02a) and the quartz-albite-muscovite band (referred as PLV-02-02b) have been analyzed separately for whole-rock geochemistry. The RMG has a strongly peraluminous composition (A/CNK = 1.13–1.45) that overlaps the fractionation trend of P-rich RMG-RMP from the FMC (Fig. 9; Tab. 1). The trace element composition shares many similarities with other RMG-RMP from the northwestern FMC, including depletion (< 25 ppm) in Sr, Sc, V, Cr, Co, Ba, Hf, Th, and Zr, and enrichment in P (1833–2357 ppm), Rb (329–1151 ppm) and Cs (58–170 ppm) compared to the upper continental crust (Fig. 10B). The rare-metal concentrations are notably elevated for Nb (65–69 ppm), Ta (46–47 ppm), W (14–37 ppm), Sn (853–3044 ppm), and Be (74–113 ppm). The REE contents are low (4.0–6.5 ppm) with a flat pattern (La/Yb = 2.9–3.6) and negative Ce and Eu anomalies (Ce/Ce* = 0.62–0.81, Eu/Eu* = 0.55–0.64), similar to other P-rich RMG-RMP from the FMC (Fig. 10D). Additionally, the studied sample has low ratios of Zr/Hf (12.7–13.2) and Nb/Ta (1.41–1.45) typical of peraluminous P-rich RMG (Ballouard et al., 2016).

Crystal chemistry of Nb–Ta–W–Ti–Sn oxides


Chemical compositions of rutile were analyzed by EMPA and LA-ICP-MS and results are reported in Supplementary Tables 2 and 3, respectively. The first type of rutile (Rt-1) corresponds to subeuhedral Fe–W-rich rutile disseminated in the tourmalinite and in quartz veins. It ranges from 250 to 1000 μm in size and shows sector zoning on BSE images (Fig. 7A). Rt-1 is characterized by core having a homogeneous composition (96.5–99.6 wt.% TiO2) and sectors having high contents of Fe (up to 1.6 wt.% FeO), Sn (up to 0.5 wt.% SnO2), and W (up to 5.0 wt.% WO3). The trace element composition shows low contents of Ta (41–250 ppm), Sc (5–29 ppm), Zn (12–63 ppm), Zr (24–170 ppm), and moderate contents of Nb (576–2463 ppm), Cu (3–369 ppm), and Cr (up to 1 wt.%). The second type of rutile (Rt-2) corresponds to Nb–Ta-rich rutile disseminated in wolframite-bearing quartz veins as grains of 50–250 μm in size (Fig. 7E). Rt-2 shows similar compositions to Rt-1 but differs by rims having high contents of Nb (up to 12 wt.% Nb2O5), Ta (up to 5.1 wt.% Ta2O5), and Fe (up to 4.8 wt.% FeO). The third type of rutile (Rt-3) consists of subeuhedral (500 μm in size) Fe–Nb-rich rutile disseminated within ferberite-1 (Fig. 7F). The chemical composition of Rt-3 is characterized by high contents of Nb (up to 14.1 wt.% Nb2O5), Ta (up to 1.3 wt.% Ta2O5), and Fe (up to 4.5 wt.% FeO), with minor amounts of W (up to 0.9 wt.% WO3) and Sn (0.6 wt.% SnO2). The last type of rutile (Rt-4) is only found in arsenopyrite as euhedral grains (ca. 250 μm) of Cr–Nb–W-rich rutile showing growth zoning (Fig. 7J). It has uncommon elevated major contents of Nb (up to 14.2 wt.% Nb2O5), Ta (up to 2.3 wt.% Ta2O5), W (up to 12.1 wt.% WO3), and Cr (up to 9.3 wt.% Cr2O3), as well as trace element enrichment of V (314–1305 ppm), Sc (64–244 ppm), Zn (10–151 ppm), Zr (100–234 ppm), and Sn (1940–2911 ppm). The different types of rutile plot along distinct chemical trends in the ternary diagram (Nb + Ta)–(Sn + Ti + W)–(Fe + Mn), reflecting variable contribution of the substitution mechanisms Ti4+−2(Fe,Mn)2++1W6++1, Ti4+−3(Fe,Mn)2++1(Nb,Ta)5++2, and Ti4+−2(Fe,Mn)3++1(Nb,Ta)5++1 (Fig. 11). Using the Zr-in-rutile thermometer of Watson et al. (2006) based on the coexistence of zircon, rutile and quartz, crystallization temperatures between 472 and 623 °C (avg = 546 ± 32 °C; n = 75) were calculated for all rutile grains (Supplementary Table 3).


Cassiterite from the RMG and from a wolframite-bearing quartz vein were analyzed by EMPA and the chemical compositions are reported in Supplementary Table 4. Magmatic cassiterite disseminated in the RMG ranges from 100 to 500 μm in size and shows a discrete zoning on BSE images (Fig. 12A). Its chemical composition is characterized by high contents of Nb (up to 2.8 wt.% Nb2O5) and Ti (up to 1.6 wt.% TiO2), and low contents of Fe (up to 0.7 wt.% FeO). Hydrothermal cassiterite hosted in a quartz-wolframite-sulfide vein consists of subeuhedral grains of 200–400 μm in size with homogeneous texture (Fig. 12B). Its chemical composition is close to the ideal formula of cassiterite with minor contents of W (up to 1.6 wt.% WO3), Fe (up to 1.5 wt.% FeO), and Nb (up to 0.2 wt.% Nb2O5). Magmatic and hydrothermal cassiterite plot along distinct geochemical trends in the ternary diagram (Nb + Ta)–(Sn + Ti + W)–(Fe + Mn), reflecting the two heterovalent substitution mechanisms Sn4+−3(Fe,Mn)2++1(Nb,Ta)5++2 and Sn4+−2(Fe,Mn)2++1W6++1 (Fig. 12C).


The W-rich variety of ixiolite (named wolframoixiolite) consists of subeuhedral grains of 200 to 500 μm in size that is found in quartz-filled veinlets cutting wolframite and as disseminations in scheelite (Figs. 7H and 7I). At the grain scale, wolframoixiolite shows a complex zoning composed of successive core, rim, and overgrowth. Chemical compositions of wolframoixiolite were determined by EMPA and are reported in Supplementary Table 5. The core and rim have overlapping compositions characterized by high contents of Nb (52.7–66.6 wt.% Nb2O5), Fe (13.6–17.1 wt.% FeO) and Mn (2.5–4.6 wt.% MnO), and low contents of W (7.0–23.1 wt.% WO3) and Ti (2.4–9.8 wt.% TiO2). In contrast, the overgrowth is distinguished by lower contents of Nb (40.4–49.7 wt.% Nb2O5) and higher contents of W (21.9–33.4 wt.% WO3). When plotted in the ternary diagram (Nb + Ta)–(Sn + Ti + W)–(Fe + Mn), the chemical composition of wolframoixiolite reflects the coupled substitution mechanisms (Nb,Ta)5+−4(Fe,Mn)2++1W6++3 and (Fe,Mn)2+−1(Nb,Ta)5+−2(Ti,Sn)4++3 (Fig. 13).


Columbite-tantalite occurs as 100–200 μm-size subeuhedral grains disseminated in the groundmass of the RMG or as microinclusions (5 to 20 μm) hosted in magmatic cassiterite. At the microscopic scale, columbite-tantalite shows a zoning composed of five distinct zones (Fig. 14A): (i) ovoid dark core of 15–20 μm in size corresponding to the initial nucleus; (ii) an irregular patchy zone close to the core; (iii) a rim-1 of medium grey color surrounding the core and showing growth banding; (iv) a rim-2 having a light grey contrast in continuity to rim-1; and (v) an overgrowth, rarely exceeding 10 μm in thickness, which shows dissolution texture at the interface with the rim and represents a late stage of crystallization. Chemical compositions of columbite-tantalite were determined by EMPA and results are shown in Supplementary Table 6. The core has higher contents of Fe (up to 11.6 wt.% FeO) and Nb (up to 65.6 wt.% Nb2O5) and lower contents of Mn (up to 9.2 wt.% MnO) and Ta (up to 19.6 wt.% Ta2O5) relative to the rims. The overgrowth shows contrasting compositions with high contents of Ta (up to 48.0 wt.% Ta2O5) and Mn (up to 10.2 wt.% MnO) and low contents of Fe (up to 7.7 wt.% FeO) and Nb (up to 36.7 wt.% Nb2O5). In the Ta/(Ta + Nb) vs. Mn/(Fe + Mn) diagram (Fig. 14B), the composition of columbite-tantalite shows a trend of increasing the Ta/(Ta + Nb) ratio (from 0.07 to 0.51) at relatively constant Mn/(Mn + Fe) ratio (0.42–0.67) from core to overgrowth. These chemical variations are related to the coupled heterovalent substitutions (Nb,Ta)5+−2(Ti,Sn)4++1W6++1 and (Fe,Mn)2+−1(Nb,Ta)5+−2(Ti,Sn)4++3, and the isovalent substitutions Nb5+−1Ta5++1 and Fe2+−1Mn2++1.

U/Pb and 40Ar/39Ar geochronology

U/Pb dating of rutile

Two samples of quartz-wolframite-arsenopyrite veins (samples PLV-7574a and PLV-02-28b) containing rutile were selected for in situ U/Pb LA-ICP-MS dating. Results are shown in Figure 15 and data are reported in Table 2. The analyzed rutile grains have low U contents ranging from 0.9 to 5.7 ppm and very low Pb contents ranging from 0.07 to 2.02 ppm. The Th contents are below 1 ppm. The isotopic compositions of rutile are variable with 238U/206Pb ratios ranging from 7.03 to 18.63 and 207Pb/206Pb ratios ranging from 0.09 to 0.54. All data are discordant in the Tera–Wasserburg diagram, which is a common feature of hydrothermal rutile (Boutin et al., 2016). Sample PLV-7574a contains hydrothermal rutile (Rt-2) disseminated in a wolframite-bearing quartz vein and is thought to be formed prior or concomitantly to wolframite deposition. The U and Pb contents of rutile range from 0.90 to 1.8 ppm and 0.07 to 0.22 ppm, respectively. The sample has relatively high radiogenic composition with 238U/206Pb values ranging from 10.93 to 18.33 and 207Pb/206Pb values from 0.11 to 0.43. The data (n = 15) define a discordia line with a lower intercept date of 322 ± 10 Ma (2σ, MSWD = 1.8; Fig. 15A). Sample PLV-02-28b contains hydrothermal rutile (Rt-4) hosted in arsenopyrite from a quartz-wolframite vein. Rutile grains from sample PLV-02-28b have low U and Pb contents ranging from 3.5 to 5.7 ppm and 0.25 to 2.0 ppm, respectively. They have variable ratios of 238U/206Pb (7.03–18.63) and 207Pb/206Pb (0.09–0.54). Linear regression of the data (n = 11) in the Tera–Wasserburg diagram yields a lower intercept date of 323 ± 5 Ma (2σ, MSWD = 1.7; Fig. 15B).

40Ar/39Ar dating of micas and adularia

Five samples containing micas and one sample with adularia were selected for in situ40Ar/39Ar dating. Results of the 40Ar/39Ar analyses are shown in Figure 16 and the dataset is reported in Table 3. Magmatic muscovite from a microgranite (sample PV-92-7) gives 40Ar/39Ar apparent ages ranging from ca. 306 to ca. 336 Ma (n = 12) and a weighted mean date of 323 ± 4 Ma (2σ, MSWD = 1.7; Fig. 16A). The 40Ar/39Ar apparent ages for magmatic biotite from a lamprophyre (sample PV-92-17) overlap between ca. 325 and ca. 347 Ma (n = 3) and yield a weighted mean date of 331 ± 14 Ma (2σ, MSWD = 0.8; Fig. 16B). Magmatic muscovite from the RMG (sample PLV-02-02b) yields 40Ar/39Ar apparent ages comprised between ca. 293 and ca. 323 Ma (n = 7) and a weighted mean date of 311 ± 5 Ma (2σ, MSWD = 2.6; Fig. 16C). Hydrothermal muscovite from a greisenized gneiss clast (sample PLV-02-01a; stage I) yields 40Ar/39Ar apparent ages ranging from ca. 308 to ca. 326 Ma (n = 6) and a weighted mean date of 318 ± 5 Ma (2σ, MSWD = 1.2; Fig. 16D). The 40Ar/39Ar apparent ages for hydrothermal muscovite from a quartz-wolframite vein (sample PV1; stage II) range between ca. 310 and ca. 338 Ma (n = 8) and yield a weighted mean date of 324 ± 5 Ma (2σ, MSWD = 1.7; Fig. 16E). Adularia from a tourmaline microbreccia (sample PLV-02-13b; stage IV) gives 40Ar/39Ar apparent ages between ca. 296 and ca. 310 Ma (n = 6) and a weighted mean date of 302 ± 4 Ma (2σ, MSWD = 1.7; Fig. 16F).

Fluid inclusions

Petrography and microthermometry of fluid inclusions hosted in quartz

Fluid inclusions (FIs) have been studied in representative quartz-wolframite vein samples (“filon 2” and “filon T18”) from the Puy-les-Vignes deposit. The studied FIs were identified in two distinct generations of quartz. Quartz Q1 forms the main gangue of the vein, similar to the breccia cement, and is interpreted as coeval with wolframite deposition (stage II). Quartz Q1 consists of large crystals having a milky color with a high density of FIs and showing evidence of ductile deformation and dynamic recrystallization. Primary zoning is locally preserved in Q1 but most FIs aligned along growth zones are decrepitated. Quartz Q2 is found in recrystallized zones of Q1 and is coeval with the base metal sulfide assemblage (stage III). In contrast to Q1, quartz Q2 corresponds to fully transparent undeformed quartz crystals containing less FIs. Based on petrography and microthermometry, three main types of quartz-hosted FIs are distinguished in both Q1 and Q2 (Tab. 4):

  • Aqueous-carbonic Lc-w: They are mainly found in Q1 either as isolated or clustered FIs, interpreted as primary, and as pseudo-secondary FI planes (FIP) sealing microcracks. Many Lc-w show reequilibration features (e.g., stretching, leaking, neonate halo) or are decrepitated (Figs. 17A17C). The Lc-w contain three phases at room temperature or upon the beginning of cooling, i.e., a carbonic liquid phase (Lcar), a carbonic vapor phase (Vcar), and an aqueous liquid phase (Laq). They occasionally contain small (< 2 μm) dark solids that could not be analyzed by Raman spectroscopy but may be graphite. The Lc-w have variable morphology, ranging from very irregular to regular (Figs. 17F17J). Their size varies between 5 and 70 μm (avg = 20 μm), with the regular FIs being generally smaller (5–20 μm) than the irregular ones (30–70 μm). The ratio between the volatile phase and the FI volume at room temperature (Rv) is highly variable (0.10–0.70, avg = 0.40). The Lc-w have low CO2 melting temperatures (Tm CO2 = −64.3 to −56.6 °C) and variable clathrate melting temperatures (Tm clat = 3.9 to 13.1 °C), indicating the presence of other gases in the volatile phase in addition to CO2 (Figs. 19A and 19B). Homogenization of the volatile phase occurs dominantly to the vapor phase (Th CO2 = −9.7 to 23.3 °C) but homogenization in the liquid or critical phase has been observed. The eutectic temperatures (Te = −24.5 to −21.1 °C) indicate an H2O–NaCl–(KCl) composition. Total homogenization of the Lc-w occurs dominantly to the liquid phase over a large range of temperatures (Th = 249 to 455 °C). Decrepitation before total homogenization of Lc-w is frequent upon heating (Td = 250 to 400 °C). Salinities calculated based on the Tm clat and Th CO2 range from 2.2 to 8.1 wt.% NaCl eq (avg = 5.3 wt.% NaCl eq) and bulk densities are comprised between 0.36 and 0.92 g cm−3 (avg = 0.67 g cm−3; Figs. 19A19C).

  • Aqueous-carbonic Lw-c: They correspond to primary and pseudo-secondary FIP forming trails along grain boundaries and growth zones in Q1 and Q2 (Figs. 17C17E). Crosscutting relations suggest that the Lw-c postdate the Lc-w. The Lw-c consist of two-phase FIs at room temperature with a dominant aqueous liquid phase (Laq) over the vapor phase (V) with variable Rv ratios (0.10–0.50, avg = 0.20). Some Lw-c hosted in Q2 contain chalcopyrite microinclusions resulting likely from heterogeneous trapping. The Lw-c have variable size (10–50 μm, avg = 20 μm) and morphology, ranging from very irregular to regular, the largest ones commonly showing irregular shapes (Figs. 17K17N). Although the carbonic liquid phase is absent, many Lw-c exhibit melting of a carbonic phase upon heating following freezing. The measured Tm CO2 varies from −103.3 to −57.4 °C, indicating the presence of other components than CO2 in the volatile phase, which is confirmed by the range of Tm clat (2.9–15.2 °C). The Lw-c hosted in Q1 and Q2 differ by their Tm CO2 and Tm clat values (Figs. 19A and 19B). The eutectic temperatures (Te = −38 to −21.1 °C) are consistent with a H2O–NaCl–(KCl) ± (MgCl2) composition. Total homogenization occurs dominantly to the liquid phase between 246 and 412 °C. Decrepitation of irregular Lw-c before total homogenization is frequent upon heating at temperatures similar to the Lc-w. Calculated salinities are comprised between 0.7 and 5.8 wt.% NaCl eq (avg = 2.6 wt.% NaCl eq) and bulk densities range from 0.47 to 0.92 g cm−3 (avg = 0.67 g cm−3; Fig. 19C).

  • Aqueous Lw: They occur as secondary FIP cutting the Lc-w and Lw-c FIs (Fig. 18D). The Lw are two-phase at room temperature with a dominant aqueous liquid phase over the vapor phase (Rv = 0.05–0.35, avg = 0.10). They are generally small (typically < 10 μm) and have regular shapes (Figs. 17O and 17P). The eutectic temperatures (Te = −23 to −21.6 °C) indicate a H2O–NaCl composition of the aqueous phase. Ice melting temperatures show a limited range of values (Tm ice = −5.5 to −0.2 °C). No clathrate melting was observed. Total homogenization always occurs to the liquid phase between 124 and 391 °C. Salinities calculated based on the Tm ice are comprised between 0.3 and 8.6 wt.% NaCl eq (avg = 4.9 wt.% NaCl eq) and bulk densities range from 0.63 to 0.99 g cm−3 (avg = 0.85 g cm−3; Fig. 19D).

Petrography and microthermometry of fluid inclusions hosted in wolframite

Wolframite from a quartz vein sample (“filon 2”) has been studied by NIR microscopy for complementing the study of FIs hosted in the coexisting quartz. Under NIR transmitted light, wolframite shows evidence of deformation and is crosscut by numerous cracks that opacify the crystal (Figs. 18A and 18B). Two main types of FIs were observed in transparent zones within wolframite: aqueous-carbonic (Lw-c) and aqueous (Lw) FIs. Three-phase volatile-rich Lc-w FIs have not been found in wolframite. The Lw correspond to secondary FIP of small (5–15 μm) two-phase (Laq + V) FIs. They have a negative crystal shape making most of them completely opaque under NIR microscopy (Fig. 18C). Only the Lw-c could be analyzed by microthermometry (Tab. 4). They correspond to primary FIs aligned along the c-axis of wolframite (Fig. 18D) and trails of pseudo-secondary FIs (Fig. 18E). The Lw-c have regular to slightly irregular shapes and their size range from 20 to 50 μm (avg = 30 μm). They are two-phase at room temperature with equal proportions of the vapor phase and the aqueous liquid phase (Rv = 0.30–0.60, avg = 0.45; Figs. 18F18M). No CO2 melting and eutectic melting temperatures were observed. Ice-melting temperatures are comprised between −5.6 and −3.0 °C and clathrate melting temperatures range from 11.0 to 14.5 °C (Fig. 19B). Total homogenization occurs to the liquid phase between 344 and 365 °C. Salinities calculated based on the Tm ice are comprised between 5.0 and 8.7 wt.% NaCl eq (avg = 6.6 wt.% NaCl eq) and bulk densities range from 0.66 to 0.73 g cm−3 (avg = 0.69 g cm−3; Fig. 19C).

Raman spectroscopy and bulk compositions of fluid inclusions

Raman spectroscopy data were acquired for quartz-hosted Lc-w and Lw-c only (Tab. 4, Figs. 19E and 19F). Due to their small size, no Raman spectroscopy analysis of Lw was possible. The Lc-w and Lw-c have a mixed volatile content composed of CO2, CH4 and N2 (Fig. 19E). The volatile phase of Lc-w is dominantly composed of CO2 (69–98 mol%) and N2 (0–26 mol%) with minor amount of CH4 (0–12 mol%) and traces of H2S (up to 0.3 mol%). In contrast, the Lw-c contain highly variable contents of CO2 (0–100 mol%), CH4 (0–86 mol%) and N2 (0–41 mol%) with traces of H2S (up to 1.1 mol%). The Lc-w and Lw-c hosted in Q1 are characterized by higher CO2/CH4 ratios relative to the ones hosted in Q2 (Fig. 19E). Compared to the Lw-c, the calculated bulk compositions of Lc-w are lower for H2O (71–88 mol%) and CH4 (0.2–4.9 mol%) but higher for CO2 (9–26 mol%). The bulk NaCl content of FIs increases (from 0.1 to 2.1 mol%) when the volatile content decreases (from 28 to 2.5 mol%; Fig. 19F). From Q1 to Q2, the Lc-w and Lw-c have increasing CH4 content in the volatile phase (from 0–20 mol% in Q1 to 20–60 mol% in Q2) while the N2 content does not vary significantly. No Raman spectroscopy measurements of FIs in wolframite could be performed due to the strong laser energy absorption by the host mineral. However, the absence of observable CO2 melting in wolframite-hosted Lw-c indicates a low volatile content while the values of Tm clat > 10 °C suggest the presence of CH4 and N2.


Timing of magmatic and hydrothermal episodes in the Puy-les-Vignes deposit

The 40Ar/39Ar mica dates obtained in this work yield a weighted mean date of 321 ± 2 Ma (2σ, MSWD = 1.7; n = 36), which falls in the same interval than the U/Pb rutile dates. Because the closure temperature for Ar diffusion in mica (ca. 350–400 °C; Harrison et al., 2009) is much lower than Pb diffusion in rutile (> 600 °C; Vry and Baker, 2006), the obtained 40Ar/39Ar mica dates correspond either to crystallization ages or to ages of isotope resetting during fluid-rock interactions. The paleodepth of formation of the Puy-les-Vignes deposit is estimated at ca. 7 km (see Sect. 9.2), which corresponds to regional temperatures of about 280 °C assuming a geothermal gradient of 40 °C km−1 typical for continental collisional orogens (Huerta et al., 1998). The estimated temperatures are lower than the temperature of Ar retention in micas implying that magmas intruded into the regional host rocks or hydrothermal fluids circulating through will cool down rapidly by thermal conduction. Therefore, we interpret the 40Ar/39Ar mica and adularia dates obtained in this work as crystallization ages.

The earliest magmatic event at Puy-les-Vignes corresponds to the emplacement of lamprophyre dykes into the Saint-Léonard-de-Noblat gneisses at the intersection with the breccia pipe. The dated lamprophyre sample yielded a weighted mean 40Ar/39Ar biotite date of 331 ± 14 Ma. The lower limit of this date is intersected by the dated microgranite that is known to crosscut the lamprophyre and yields a weighted mean 40Ar/39Ar muscovite date of 323 ± 4 Ma. This indicates an episode of mantle melting between ca. 345 and ca. 327 Ma, similar to vaugnerite dykes dated at 334 ± 1 Ma and 336 ± 2 Ma in the eastern FMC (Laurent et al., 2017). The date of the microgranite falls into the age range of two-mica peraluminous granitoids from the northwestern FMC (Fig. 20), such as the Saint-Sylvestre leucogranitic complex (324 ± 4 Ma; Holliger et al., 1986) or the nearby Auriat granite (324 ± 1 Ma; Gebauer et al., 1981). Therefore, we interpret the 40Ar/39Ar muscovite date for the microgranite to represent the crystallization age. The last known magmatic event at Puy-les-Vignes corresponds to the intrusion of a RMG that yielded a weighted mean 40Ar/39Ar muscovite date of 311 ± 5 Ma. This date falls in the same age interval of ca. 315–310 Ma than other RMG-RMP in the Limousin (Fig. 20; Cheilletz et al., 1992; Alexandrov et al., 2000; Cuney et al., 2002; Melleton et al., 2015). Because the studied RMG lacks hydrothermal alteration features, the 40Ar/39Ar date of 311 ± 5 Ma is interpreted as the crystallization age.

Based on field relationships and petrographic observations, the Puy-les-Vignes breccia pipe results from a multistage hydrothermal evolution. The 40Ar/39Ar dates of 324 ± 5 Ma and 318 ± 5 Ma on hydrothermal muscovite and the U/Pb dates on rutile of 322 ± 10 Ma and 323 ± 5 Ma from stages I and II overlap between ca. 323 and ca. 318 Ma (Fig. 20). These ages are consistent with the 40Ar/39Ar plateau age of 323 ± 2 Ma obtained on a muscovite interpreted to be prior or coeval to wolframite deposition (Cuney et al., 2002). U/Pb dating of wolframite yielded a concordia age of 317.7 ± 0.7 Ma (Harlaux et al., 2018a), which is compatible with the previous ages and points toward a W mineralization episode at ca. 318 Ma. The period of ca. 325 to ca. 315 Ma is synchronous to the emplacement of two-mica peraluminous leucogranites in the Limousin such as the La Porcherie granite (317 ± 3 Ma; Lafon and Respaut, 1988), the Courcelles granite (318 ± 3 Ma; Cartannaz et al., 2007), the Glény granite (318 ± 3 Ma; Faure et al., 2009b), the Pradines granite (313 ± 4 Ma; Gébelin et al., 2009), and the Livradois granite (315 ± 4 Ma; Gardien et al., 2011). Crustal melting was coeval with high-temperature and low-pressure metamorphism as evidenced by cordierite-sillimanite migmatites that formed at 315 ± 5 Ma and 316 ± 2 Ma along the western border of the Millevaches leucogranitic complex (Gébelin et al., 2009). The 40Ar/39Ar adularia date of 302 ± 4 Ma (stage IV) records a late hydrothermal episode that is temporally disconnected from previous stages. This age is synchronous with the cooling and exhumation of the Limousin metamorphic basement at ca. 305–300 Ma (Scaillet et al., 1996a; Le Carlier de Veslud et al., 2013) related to the late- to post-orogenic uplift and generalized extension of the FMC (Faure et al., 2009a). The Bi ± Au–Ag assemblage (stage V) was therefore formed during or after the 305–300 Ma period coevally with late Carboniferous orogenic gold deposits in the FMC, which share similar mineralogical features and result from the mixing of metamorphic and meteoric fluids (Boiron et al., 2003; Bouchot et al., 2005).

Nature and pressure-temperature conditions of the mineralizing fluids

The W-mineralizing fluids in the Puy-les-Vignes deposit are characterized by a low to moderate salinity (0.7–8.7 wt.% NaCl eq), elevated homogenization temperatures (250–455 °C), and high contents of CO2 (avg = 10 mol%), CH4 (avg = 1.6 mol%), and N2 (avg = 1.1 mol%). These fluid compositions belong to the C–O–H–N system and are similar to other peri-granitic W–Sn deposits in the FMC (Ramboz et al., 1985; Bril and Beaufort, 1989; Cuney et al., 1992; Lerouge et al., 2000; Vallance et al., 2001). Similar C–O–H–N fluids have been described in other Variscan W–Sn deposits (Noronha et al., 1992; Vindel et al., 1995; Chicharro et al., 2016), in evolved pegmatites and leucogranites (Charoy and Noronha, 1996; Fuertes-Fuente et al., 2000; Bakker and Schilli, 2016), and in metamorphic environments (Van den Kerkhof et al., 1991; Guedes et al., 2002; Boiron et al., 2007). The presence of CO2 and CH4 in FIs can result from different mechanisms including high-temperature graphite-water interactions in metasediments (Ramboz et al., 1985; Dubessy et al., 1989), degassing or flushing of CO2 from silicate melts (Lowenstern, 2001; Blundy et al., 2010), CO2 reduction by hydrogen (Hall and Bodnar, 1990), or boron complexing in aqueous-carbonic fluid (Huff and Nabelek, 2007). The release of N2 can originate from the breakdown of NH4+-bearing mica and feldspar during dehydration-melting of metasediments (Moine et al., 1994; Bebout et al., 1999; Pitcairn et al., 2005) or from strongly reduced silicate melts (Boulliung et al., 2020).

Raman spectroscopy data of FIs from Puy-les-Vignes show an evolutionary trend between a CO2 (> CH4)-rich fluid endmember and a CH4 (> CO2)-rich fluid endmember from Q1 to Q2 (Fig. 19E), which is compatible with decreasing temperatures of the graphite-water equilibrium from > 550 °C to 400 °C (Huizenga, 2001). While the relatively high N2 content of FIs could indicate a metamorphic component in the hydrothermal system, the variable CH4 to N2 ratios suggest a two-component origin. Therefore, we interpret the C–O–H–N fluid compositions as resulting from a moderate-salinity (> 9 wt.% NaCl eq) and high-temperature (> 450 °C) hydrothermal fluid that interacted with volatile-rich metasedimentary rocks and/or that mixed with low-salinity metamorphic fluids equilibrated with such rocks. The Lw show a trend of decreasing salinity (from > 8.0 to < 0.5 wt.% NaCl eq) with decreasing temperature (from > 350 to < 150 °C), which is interpreted as a cooling-dilution trend (Fig. 19D). The Th variations for a given composition are interpreted as recording anisothermal mixing under fluctuating pressure conditions from lithostatic to hydrostatic (Dubessy et al., 2003) and partially disturbed by post-entrapment reequilibration (e.g., necking-down). Two fluid endmembers are involved in the mixing, i.e. a high-temperature (> 350 °C) and moderate-salinity (> 9 wt.% NaCl eq) fluid L1, similar to the mineralizing fluids from the stages II and III, and a low-temperature (< 150 °C) and low-salinity (< 0.5 wt.% NaCl eq) fluid L2, possibly of meteoric origin.

Pressure-temperature conditions of the Puy-les-Vignes deposit were reconstructed based on representative isochores (Fig. 21). No boiling assemblages were observed in the FI inventory, which precludes determining directly entrapment pressures based on homogenization temperatures. In Q1, two families of isochores are distinguished for the Lc-w and Lw-c (Fig. 21A): (i) high-angle isochores corresponding mostly to the regular and small FIs; and (ii) low-angle isochores corresponding to the majority of irregular and large FIs showing reequilibration features and to large and regular FIs. The two sets of isochores intersect geothermal gradients at 60–70 °C km−1 for trapping temperatures of ca. 400–450 °C and pressure variations from ca. 190 MPa (lithostatic) to ca. 70 MPa (hydrostatic). These pressure estimates indicate that the Puy-les-Vignes deposit formed at a paleodepth of ca. 7 km. Similar trapping conditions are estimated for the Lw-c hosted in wolframite. The pressure variations, from lithostatic to hydrostatic, are interpreted to result from the hydraulic fracturing of the host rocks during the breccia pipe formation. The differential pressure of 120 MPa between the lithostatic and hydrostatic gradients is equal to the minimum ΔP required to reequilibrate a 20 μm diameter FI hosted in quartz (Bodnar et al., 1989). Trapping temperatures are more than 100 °C lower than those calculated by the Zr-in-rutile thermometer of 546 ± 32 °C, suggesting that hydrothermal rutile yields overestimated temperatures, as also documented by Cabral et al. (2015). Representative isochores for the Lc-w and Lw-c hosted in Q2 yield similar trapping conditions of ca. 400–420 °C and pressure fluctuations from ca. 190 MPa (lithostatic) to ca. 70 MPa (hydrostatic) for high geothermal gradients of 60–70 °C km−1 (Fig. 21B). Pressure variations from ca. 135 MPa (lithostatic) to ca. 50 MPa (hydrostatic) are estimated for trapping of the Lw, when the same ΔP is encountered for the L1 (380 °C with a 70 °C km−1 gradient) and L2 (210 °C with a 40 °C km−1 gradient) fluid endmembers (Fig. 21C). This would correspond to a paleodepth of ca. 5 km for trapping of Lw. Compared to previous stages, the Lw records a decrease of the reservoir depth (from 7 to 5 km) likely related to an increased exhumation. Assuming an erosion rate of 0.5–1.0 mm yr−1, representative for the exhumation of the Variscan basement during the late Carboniferous (Scaillet et al., 1996b; Schulmann et al., 2002), the erosion of 2 km of rock thickness between the stage II and the Lw stage would require between 2 and 4 My. Considering that the W mineralization formed at ca. 318 Ma, the Lw would have been trapped between 314 and 316 Ma. The Lw represent, therefore, a late stage in the Puy-les-Vignes hydrothermal system, which possibly records the infiltration of meteoric waters during the late Carboniferous uplift of the FMC, as also evidenced in other W–Sn deposits (Vallance et al., 2001) and orogenic Au deposits (Boiron et al., 2003) from the Limousin area.

Hydrothermal fluid origin and wolframite precipitation

The Puy-les-Vignes breccia pipe resulted from intense hydrofracturing of the Saint-Léonard-de-Noblat gneiss unit caused by ascending overpressured hydrothermal fluids, as evidenced by characteristic textures of hydraulic brecciation and fluid-assisted rotation of metric-size clasts (Jébrak, 1997). Wolframite from Puy-les-Vignes shows similar trace element composition than evolved peraluminous granites from the FMC (Harlaux et al., 2018b), suggesting that W and other metals were likely sourced from an unexposed leucogranite located beneath the breccia pipe. Tungsten could be released either by magmatic fluids due to elevated fluid-melt-crystal partitioning coefficients (Hulsbosch et al., 2016; Schmidt et al., 2020) or by metamorphic fluids during high-temperature interactions with a two-mica leucogranite (Vallance et al., 2001; El Korh et al., 2020; Carocci et al., 2021). The role of fluid-rock interactions is supported by the chemical and boron isotope compositions of tourmaline that are compatible with both a magmatic and a metamorphic fluid source (Harlaux et al., 2019) as well as by the Fe-dominated composition of wolframite (Michaud and Pichavant, 2019). The origin of the hydrothermal fluids (magmatic vs. metamorphic) can be further discussed based on estimation of the volume of fluids required for precipitation of quartz and wolframite (Supplementary Table 7). Considering an elliptic surface of dimensions 340 × 80 m and a vertical extension of 600 m (i.e., twice the known height) for the breccia pipe, the total volume of rock would be of 5.1 × 107 m3. Assuming that quartz represents about 10% of the total volume of the breccia (Weppe, 1951, 1958), a mass of 1.3 × 1010 kg of quartz can be estimated. Taking into account a quartz solubility of 55 mmol kg−1 for a hydrothermal fluid at 10 wt.% NaCl eq, 500 °C and 100 MPa (Monecke et al., 2018) and an average fluid density of 0.67 g cm−3, we estimate a volume of fluids of 6.1 × 109 m3 (6.1 km3). Following the same reasoning for wolframite, we can estimate a volume of fluids of 1.6 × 108 m3 (0.16 km3) by assuming an average grade of 0.5% WO3 and considering a W solubility of 500 ppm for a fluid at 10 wt.% NaCl eq, 500 °C, and 100 MPa (Wood and Samson, 2000). Therefore, the minimum volume of hydrothermal fluids required for the formation of the wolframite-bearing breccia pipe would be of about 6.2 km3. It is noteworthy that these calculations assume 100% efficiency of mineral precipitation and no fluid loss in the host rocks, implying that the volume of fluids involved should be larger than this estimation.

  • Magmatic fluid origin: Assuming that the volume of fluids of 6.2 km3 was entirely produced by a water-saturated intrusion containing an initial water content of 5 wt.%, this would require a minimum volume of melt of ca. 36 km3 for an average melt density of 2.3 g cm−3 (Supplementary Table 7). Considering a volume contraction of 10% from liquidus to solidus (Bea, 2010), this would correspond to a volume of granite of 33 km3, i.e., a pluton of dimensions 1.5 × 4.7 × 4.7 km. Fluids that exsolve from an intrusion at a depth > 6 km (corresponding to a lithostatic pressure > 150 MPa) are in the single-phase domain at magmatic conditions and remain in this state during cooling to 500 °C (Driesner and Heinrich, 2007). Consequently, fluids trapped at these pressure-temperature conditions have a moderate salinity (ca. 10 wt.% NaCl eq), an intermediate density (ca. 0.6 g cm−3) and do not undergo boiling. This model explains the absence of coexisting vapor and hypersaline fluid inclusions (boiling assemblage), the moderate fluid salinity (< 9 wt.% NaCl eq), as well as the high geothermal gradients of 60–70 °C km−1 determined from the P–T reconstruction. The overpressures generated by the exsolved magmatic fluids could be a mechanism for the hydraulic brecciation of the host rocks, as generally proposed for the formation of breccia pipes (Burnham, 1985; Sillitoe, 1985; Jébrak, 1997). The high CO2 content (avg = 10 mol%) determined in FIs fall in the same range than magmatic fluids from porphyry systems (Kouzmanov and Pokrovski, 2012). In this scenario, magmatic fluids exsolved from an evolved leucogranite would have interacted with volatile-rich metamorphic rocks resulting in the cooling, dilution, and enrichment in metamorphic-derived volatiles (mainly CH4 and N2). The main limitation of this model is the volume of fluids required since numerical modelling shows that magmatic fluids are released by successive pulses from crystallizing intrusions (Weis et al., 2012; Lamy-Chappuis et al., 2020). This would imply a much larger volume of granitic melt that the one estimated in order to supply in a single time the volume of fluids required to generate the breccia pipe.

  • Metamorphic fluid origin: Considering that the fluids were produced by devolatilization reactions of volatile-rich metasedimentary rocks, the question arises about the source reservoir. The LGU hosting the Puy-les-Vignes deposit experienced migmatization during the Middle Devonian (385–375 Ma) at metamorphic conditions of 650–750 °C and 700 MPa (“D1 event” of Faure et al., 2009a) implying the muscovite- and biotite-breakdown reactions. Consequently, the LGU paragneisses cannot be a local source of metamorphic fluids at 318 Ma. In contrast, the PAU, structurally located below the LGU, never experienced the D1 metamorphism nor migmatization and records maximum metamorphic conditions of 500–600 °C and 700–800 MPa (Do Couto et al., 2016) during the early Carboniferous (360–350 Ma; “D2 event” of Faure et al., 2009a). Thus, the PAU micaschists underlying the LGU gneisses may represent a distal source of metamorphic fluids produced by high-temperature and low-pressure metamorphism during the late Carboniferous (320–315 Ma). Deep metamorphic fluids could be gathered into a collection zone, then focused and channelized in a mid-crustal fault zone, ascending at high rates up to a dispersion region in cooler rocks (Thompson and Connolly, 1992; Connolly, 1997; Carocci et al., 2021). The half-width of both collection and dispersion zones is typically in the 3–5 km range, and the height of the system is on the order of 10 to 15 km. This model accounts for the high volume of fluids required, the upward heat advection, and the high CO2–CH4–N2 contents of the fluids. The rapid and adiabatic ascending of pressurized volatile-rich fluids along a crustal fault zone under the LGU carapace may represent a mechanism for generating hydraulic fracturing of the host rocks. The main limitation of this model is that devolatilization of graphite-rich pelitic metasediments only produce low-salinity (typically < 5 wt.% NaCl eq) metamorphic fluids (Guedes et al., 2002; Yardley and Graham, 2002).

Based on available data, we favor an intermediate model involving a W-rich intermediate-density magmatic fluid that exsolved from an evolved leucogranite and interacted with the volatile-rich metasedimentary country rocks and/or possibly mixed with low-salinity metamorphic fluids of deep origin (Fig. 22). Experimental works and thermodynamic modelling show that the solubility of wolframite in hydrothermal fluids increases with increasing temperature and salinity and with decreasing pH and Fe contents of the fluid (Wood and Samson, 2000; Wang et al., 2019). Addition of non-polar molecules such as N2, CO2, and CH4 yields a decrease of W concentration in solution (Dubessy et al., 1989; Gibert et al., 1992). Precipitation of wolframite at Puy-les-Vignes could have been triggered by several processes including cooling, depressurization, fluid mixing, and fluid-rock interaction (Heinrich, 1990; Wei et al., 2012; Dewaele et al., 2016; Chicharro et al., 2016; Lecumberri-Sanchez et al., 2017; Legros et al., 2019, 2020; Monnier et al., 2020). In particular, fluid pressure drop induced by hydrofracturing of the host rocks results in significant decrease of CO2 solubility and pH increase, which can cause in turn wolframite precipitation (Liu et al., 2017).

Overprint of the W-mineralized system by a rare-metal granite

Wolframite deposition in the Puy-les-Vignes deposit occurred during the main ore stage II at 318 Ma and was followed by precipitation of accessory W–Nb–Ta oxide minerals (wolframoixiolite, Ta–Nb–ferberite, and W–Nb–Ta–rutile). Wolframoixiolite is found as infill of dissolution microvugs in wolframite, in quartz-filled veinlets cutting wolframite, and as disseminations overgrown by ferberite-1 within scheelite; all together suggesting that the fluid having precipitated wolframoixiolite was out of equilibrium with wolframite. Ferberite-1 cuts and replaces wolframite along crystal borders and in dissolution microvugs and locally shows Ta–Nb-rich primary zoning. Such primary textures are similar to those described for Nb-rich ferberite from the Echassières W deposit in the northern FMC (Aïssa et al., 1987; Harlaux et al., 2018a, b). Nb–Ta-rich rutile is found as disseminations within ferberite-1 and forms overgrowths on Fe–W–rutile, the latter being interpreted as coeval with wolframite deposition similarly to the findings of Carocci et al. (2019) in the Panasqueira W deposit (Portugal). Several studies showed that the association of Ta–Nb–ferberite, W–Nb–Ta–rutile, wolframoixiolite, and columbite-tantalite minerals is a marker of the magmatic-hydrothermal transition in RMG-RMP (Breiter et al., 2007, 2017; Alekseev et al., 2011; René and Škoda, 2011; Zhu et al., 2015, 2019; Michaud et al., 2020). The studied sample of RMG exhibits all the mineralogical and geochemical features of a highly fractionated P-rich peraluminous RMG cupola (Cuney et al., 1992; Raimbault et al., 1995; Zhu et al., 2001; Breiter et al., 2005; Linnen and Cuney, 2005) and shows similarities with other P-rich peraluminous RMG-RMP from the northern FMC (Aubert, 1969; Mourey, 1985; Cuney et al., 1992; Raimbault et al., 1995; Raimbault, 1998). Additionally, columbite-tantalite disseminated in the RMG exhibits an increasing Ta/(Ta + Nb) ratio at relatively constant Mn/(Mn + Fe) ratio from core to rim, as also documented in other RMG-RMP from the northern FMC (Wang et al., 1992; Belkasmi and Cuney, 1998; Belkasmi et al., 2000). The progressive increase in the Ta/(Ta + Nb) ratio may reflect either a magmatic fractionation trend due to the higher solubility of tantalite relative to columbite in peraluminous melts, or the effect of fluid-mineral interaction during a late hydrothermal remobilization (Linnen and Cuney, 2005; Van Lichtervelde et al., 2007; Breiter et al., 2007; Zhu et al., 2015; Wu et al., 2018).

Experimental works and fluid-melt-crystal partitioning calculations show that W partitions readily into magmatic fluid relative to the silicate melt in peraluminous granites (Zajacz et al., 2008; Hulsbosch et al., 2016; Schmidt et al., 2020). In contrast, Nb and Ta preferentially partition into the silicate melt (Linnen and Cuney, 2005; Borodulin et al., 2009). However, the solubility of Nb and Ta is strongly enhanced in F-rich hydrothermal fluids as shown by experimental studies (Zaraisky et al., 2010; Timofeev et al., 2015, 2017). As a consequence, F-rich magmatic fluids released during crystallization and water saturation of a RMG melt can transport significant amounts of W, Nb, and Ta. This is confirmed by elevated concentrations of W (> 100 s ppm) and Nb–Ta (> 1000 s ppm) measured in individual primary fluid inclusions (400 to > 600 °C, 18 to 48 wt.% NaCl eq) from the Beauvoir RMG, for which a F content of about 1.6–2.0 wt.% was estimated in the magmatic fluid (Harlaux et al., 2017). The nearby La Bosse quartz-ferberite stockwork was strongly overprinted by the magmatic fluids issued from the Beauvoir granite, resulting in crystallization of topaz, partial dissolution of quartz, and precipitation of accessory Li-micas and W–Nb–Ta oxides (Aïssa et al., 1987; Cuney et al., 1992). Specifically, Ta-rich ferberite, wolframoixiolite, and columbite-tantalite minerals were described in dissolution microvugs within ferberite and were interpreted to be precipitated by the magmatic fluids released from the Beauvoir granitic melt (Cuney et al., 1992; Harlaux et al., 2017). Consequently, we propose that the W–Nb–Ta oxide assemblages observed in the quartz-wolframite veins from the Puy-les-Vignes deposit were precipitated from a magmatic-hydrothermal fluid issued from a concealed P-rich peraluminous RMG that overprinted the W-mineralized system (Fig. 22).


We showed that the formation of the Puy-les-Vignes W breccia pipe results from a multistage development during 25 million years related to four major episodes in the late Carboniferous. The first episode corresponds to the emplacement of an unexposed peraluminous granite at ca. 324 Ma, which generated microgranite dykes exposed at the present-day surface. The second episode resulted in the formation of the quartz-supported breccia pipe and wolframite mineralization at ca. 318 Ma at a paleodepth of ca. 7 km. The mineralizing fluids have a H2O–NaCl–CO2–CH4–N2 composition, a moderate-salinity (< 9 wt.% NaCl eq) and were trapped at high-temperatures (> 400 °C) during lithostatic to hydrostatic pressure variations caused by hydrofracturing of the host rocks. We interpret wolframite precipitation as the result of a W-rich intermediate-density magmatic fluid that exsolved from an evolved leucogranite and interacted with volatile-rich metasedimentary country rocks and/or possibly mixed with low-salinity metamorphic fluids of deep origin. The third episode corresponds to a magmatic-hydrothermal Nb–Ta mineralization interpreted to be related to the intrusion at ca. 311 Ma of a P-rich peraluminous RMG overprinting the W-mineralized system. The last episode corresponds to a disseminated Bi ± Au–Ag mineralization emplaced at ca. 300 Ma, which shares mineralogical features with late Carboniferous orogenic gold deposits in the FMC.


This work was supported by the French National Research Agency through the national program “Investissements d’avenir” of the Labex Ressources 21 (ANR-10-LABX-21-RESSOURCES21) and the ERAMIN project NewOres (ANR-14-EMIN-0001). The authors greatly thank Dr. Hossein Ahmadzadeh and the Musée Minéralogique de l’École des Mines d’Alès (France) for supplying samples of collection from the Puy-les-Vignes mine. We also thank Drs. Hélène Legros and Patrice Bruneton for their assistance in the field. Technical support by Sandrine Mathieu, Dr. Andreï Lecomte (SEM), Dr. Olivier Rouer (EMPA), Dr. Chantal Peiffert (LA-ICP-MS) and Dr. Marie-Camille Caumon (Raman spectroscopy) is gratefully acknowledged. We acknowledge the GeOHeLiS platform (Univ. Rennes, France) for the LA-ICP-MS U–Th/Pb dating facilities. Drs. Vincent Casanova and Bertrand Rottier are also warmly thanked for their help during the infrared microthermometric analyses of wolframite at the University of Geneva, Switzerland. Finally, we would like to thank an anonymous reviewer and the Guest Editor Dr. Eric Marcoux for their critical comments that helped to improve the original manuscript.

Cite this article as: Harlaux M, Marignac C, Mercadier J, Poujol M, Boiron M-C, Kouzmanov K, Camacho A, Alikouss S, Roméo B, Mouthier B, Cuney M. 2021. Multistage development of a hydrothermal W deposit during the Variscan late-orogenic evolution: the Puy-les-Vignes breccia pipe (Massif Central, France), BSGF - Earth Sciences Bulletin 192: 33.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.