Late Ediacaran juvenile magmatism in the Variscan Monts-du-Lyonnais metamorphic complex (Massif Central, France)

Within the framework of plate tectonics theory, sutures, i.e. zones along which oceanic lithospheres were subducted, are key to deciphering the evolution of orogens (Dewey, 1987). Their identification remains challenging in the case of old and deeply eroded collisional orogenic systems such as the European Variscan belt, formed through the convergence of Laurussia and Gondwana during the late Paleozoic (Edel et al., 2018; Matte, 1986; Stampfli et al., 2013). Indeed, the diagnostic suture rock associations (high-pressure metamorphic belts and ophiolite sequences) have been largely dismantled by early erosion and intensely reworked by collisional processes, leading to conflicting views on the number of suture zones and their associated oceans (see discussions in Franke et al., 2017; Kroner and Romer, 2013). Paleomagnetic data and paleobiogeography (Domeier, 2016; Paris and Robardet, 1990; Van der Voo et al., 1980) suggest that the formation of the Variscan belt followed the closure of a single >1000 km-wide Ordovician oceanic domain called the Rheic Ocean. On the other hand, petrological evidence coupled with geophysical data collectively indicate the existence of at least three sutures (Fig. 1a, Schulmann et al., 2022). The subduction of an Ordovician-early Devonian “Mid-Variscan” ocean (Galicia/Southern Brittany/Tepla ocean of Matte, 1986) would have initiated in the mid- to late-Devonian period (390–360 Ma). The narrower Saxothuringian ocean would have closed shortly after in the early Carboniferous (360–340 Ma). Lastly, subduction of the Devonian Lizard–Rhenohercynian Ocean would have occurred at 340–330 Ma (Zeh and Gerdes, 2010).

Delineating the suture zones along the belt is of marked importance to unravel the architecture of the European continental crust and the evolution of the Variscan orogeny. However, accurately correlating their locations and geometries in the eastern French Massif Central (eFMC, Fig. 1a) has proven challenging due to a lack of comprehensive seismic survey and the absence of ophiolite sequences (Schulmann et al., 2022). In the eFMC, it is commonly inferred (Lardeaux et al., 2014; Matte, 1986; Pin, 1990; Vanderhaeghe et al., 2020) that the Mid-Variscan suture is stamped by distinctive heterogeneous lithological formations referred to as “Leptynite–Amphibolite Complexes” (LACs, see review in Santallier et al., 1988). The latter encompass mafic meta-igneous rocks occurring as metric to kilometric boudins embedded in paragneisses and/or are intimately associated with metarhyolites/metagranites of crustal derivation (Briand et al., 1995, 1991; Chelle-Michou et al., 2017; Pin and Marini, 1993). The metabasites often bear relics of Devonian (385–360 Ma) high-pressure metamorphism defining a P–T gradient consistent with subduction (Lardeaux, 2023). Ultramafic bodies representing subducted mantle material are also present (Gardien et al., 1990). Early geochronological studies suggested that both mafic and felsic igneous protoliths were emplaced contemporaneously during the Cambrian–Ordovician period (Pin and Lancelot, 1982). This has led many authors to interpret the wide variety of geochemical signatures displayed by mafic rocks, ranging from MORB-type tholeiites to back-arc basalts and OIB (Briand et al., 1995, 1991; Downes et al., 1989; Giraud et al., 1984; Piboule and Briand, 1985; Pin and Marini, 1993) as reflecting an emplacement in an attenuated continental lithosphere at the incipient stages of oceanization (Lardeaux et al., 2014 and references therein). In this context, the LACs were proposed to represent fragments of the passive margin to ocean–continent transition zone of the Mid-Variscan ocean, which were subsequently buried through oceanic subduction and later exhumed as tectonic mélanges along major thrust zones in the orogenic wedge (Burg and Matte, 1978; Lardeaux et al., 2014, 2001), or possibly extruded through the upper plate (de Hoÿm de Marien et al., 2023; Keppie et al., 2010; Maierová et al., 2021).

Following the recognition of the main nappe architecture in the eFMC during the mid-1980s, the various LACs were grouped into a single litho-tectonic unit referred to as “the LAC” (singular) and were assumed to share the same geodynamic significance (Bouchez and Jover, 1986; Lardeaux et al., 2014; Matte, 1986; Vanderhaeghe et al., 2020). This broad correlation was questioned by Santallier et al. (1988), who pointed out the lack of age data on the protoliths and the uncertainty on the respective petrogenesis of individual LACs. This concern is echoed in the adjacent Maures Massif, where three LACs were initially identified, yet only two are currently regarded as delineating a suture zone (Bellot et al., 2010; Briand et al., 2002; Jouffray et al., 2023; Schneider et al., 2014). Therefore, a reexamination of individual (ultra)mafic–felsic rock associations across the eFMC appears necessary to ascertain whether each LAC record the same geodynamic events and collectively represent a single litho-tectonic unit. In this context, the increasing versatility of in situ laser ablation–inductively coupled plasma–mass spectrometry (LA–ICP–MS) zircon U–Pb dating offers new opportunities to unravel the protolith ages of the LAC rocks and clarify their origin (Chelle-Michou et al., 2017; de Hoÿm de Marien et al., 2023; Lotout et al., 2020; Paquette et al., 2017).

This contribution focuses on the Riverie LAC from the Monts-du-Lyonnais metamorphic complex in the eFMC. Novel petrological and geochronological data demonstrate that this mafic–felsic association consists of juvenile arc/back-arc-derived magmatic rocks of late Ediacaran age which are unrelated to the opening and closure of the Mid-Variscan ocean. These findings highlight the necessity for a reevaluation of the tectonic history of individual LACs within the eFMC.

The eFMC is part of the internal zone of the Variscan belt (Fig. 1), primarily composed of metamorphic nappes intruded by various Carboniferous granitoids (see review in Faure et al., 2009). In the northern part of the eFMC, metamorphic rocks are overlain by Visean sediments and volcanics indicating that the main orogenic phase occurred before 345 Ma. In contrast, compression persisted till the Namurian (corresponding to the Serpukhovian and Bashkirian pro parte, c. 325 Ma) in the foreland basin to the south. Upper Carboniferous (Stephanian, i.e. Kasimovian–Gzhelian pro parte) to Permian sediments were unconformably deposited over the metamorphic-plutonic basement in a late to post-orogenic setting. The nappe pile consists of the high-grade “Upper Gneiss Unit” overlying a crustal package of parautochtonous terrains, sometimes divided into the Lower Gneiss Unit and the (lower-grade) Parautochtonous Unit (Ledru et al., 1989). Southwards-directed thrusting of the Upper Gneiss Unit was coeval to the development of an inverted Barrovian metamorphic sequence (Burg et al., 1984) at 360–345 Ma (Chelle-Michou et al., 2017; Costa, 1989). The LACs are commonly located at the base of the Upper Gneiss Unit and mark the contact between this unit and the underlying parautochtonous terrains. Within most LACs, mafic meta-igneous rocks preserve a record of high-pressure subduction-related metamorphism with pressures exceeding 15 kbar at 385–360 Ma (de Hoÿm de Marien et al., 2023; Joanny et al., 1991; Lardeaux et al., 2001; Lotout et al., 2020, 2018). The parautochtonous terrains were locally affected by a low-pressure high-temperature Buchan-type metamorphic event coeval with crust extension and the formation of the Velay and Montagne Noire gneiss domes (Gardien et al., 2022, 1997; Ledru et al., 2001). The latter hosts mafic eclogites whose origin is disputed: they could either correspond to Upper Gneiss Unit material (and thus be subduction-related, Pitra et al., 2022) or lower crustal rocks exhumed during doming (Whitney et al., 2020). Lastly, the Brévenne unit is a bimodal volcanic sequence of late Devonian age metamorphosed under greenschist- to lower amphibolite-facies conditions and juxtaposed to the Upper Gneiss Unit during the main collisional phase (Feybesse et al., 1995).

The protoliths of the Upper Gneiss Unit and parautochtonous terrains consist of Ediacaran to Ordovician detrital (volcano-)sedimentary units, intruded by several generations of plutonic rocks (Couzinié et al., 2022, 2019). No rock older than Neoproterozoic has been described so far. Pre-Variscan magmatic rocks include: (i) latest Ediacaran (c. 545 Ma) and Cambro-Ordovician granites and rhyodacites of crustal origin, found throughout the nappe pile regardless the metamorphic grade (Chelle-Michou et al., 2017; Couzinié et al., 2022, 2017; Couzinié and Laurent, 2021); (ii) mafic (gabbroic and basaltic) rocks, mainly in the LACs of the Upper Gneiss Unit but also in the parautochtonous terrains as lava flows, pyroclastic deposits, and subvolcanic sills (Briand et al., 1995, 1992). Mafic rocks exhibit a large range of geochemical signatures ranging from tholeiitic to alkaline and calc-alkaline (Briand et al., 1992; Pouclet et al., 2017). The crust-derived magmas attest to the reworking of Ediacaran sediments (Couzinié et al., 2022, 2017), while mafic magmas would originate from the lithospheric and asthenospheric mantle, with the subordinate involvement of a garnet-bearing source (Pin and Marini, 1993; Pouclet et al., 2017). The latest Ediacaran crustal melting was possibly a distal manifestation of the Cadomian orogenic events (Couzinié et al., 2017), while Cambro-Ordovician crust-mantle magmatism should be tied to protracted lithosphere extension that prevailed at that time along the northern Gondwana margin (Chelle-Michou et al., 2017; Couzinié et al., 2022; Pin, 1990; Pin and Marini, 1993).

The Monts-du-Lyonnais metamorphic complex (Fig. 2) belongs to the Upper Gneiss Unit. Its litho-tectonic pile consists of a set of amphibolite-(granulite-)facies metamorphic rocks including from bottom to top (Feybesse et al., 1995): (i) a lower metasedimentary unit corresponding to cordierite-bearing migmatites and garnet–sillimanite–biotite paragneisses in the northwestern and southeastern part of the complex, respectively, (ii) three (ultra)mafic–felsic rock associations cartographically corresponding to “belts” and assimilated to LACs by Chantraine et al. (2003); (iii) mildly peraluminous, locally anatectic orthogneisses; (iv) an upper metasedimentary unit, also migmatitic, with relics of high-pressure granulite-facies metamorphism. The first, southernmost (ultra)mafic–felsic belt includes 100–3000 m-long amphibolite boudins corresponding to retrogressed eclogites stretched within a paragneiss matrix and associated with garnet-bearing mantle peridotites (Blanc, 1981; Gardien et al., 1990). The second, northernmost Chaussan belt is composed of fine-grained leucogneisses (leptynites) enclosing numerous hecto- to kilometric massifs of meta-igneous mafic and felsic granulites (Dufour, 1985). Finally, the Riverie belt located in-between the previous two features an association of amphibolites and amphibole-bearing gneisses, locally leucocratic (leptynites) (Blanc, 1981).

Metasediments experienced peak metamorphic conditions of 750–800 °C at 7–9 kbar coeval to the development of the regional foliation S_1-2 (Feybesse et al., 1995; Gardien et al., 1990). Anatexis initiated at that stage and persisted during exhumation down to c. 4 kbar. The age of this anatexis is poorly constrained, as the only published data consist of a whole-rock Rb–Sr isochron date of 384 ± 16 Ma, interpreted as the age of partial melting (Duthou et al., 1994). Relictual kyanite in metasediments suggests an early high-pressure evolution at P > 10 kbar (Feybesse et al., 1995). Rocks from the LACs record different metamorphic conditions. Eclogites from the southern belt experienced peak conditions of P > 15 kbar and 750–850 °C (Dufour et al., 1985; Joanny et al., 1991). One coesite grain retrieved from an eclogite sample would even suggest P > 28 kbar (Lardeaux et al., 2001). Meta-igneous granulites from the Chaussan belt record P of 8–10 kbar and T in the range 800–870 °C (Dufour, 1985). Metamorphic conditions for the Riverie belt have not been estimated but Feybesse et al. (1995) stressed that these rocks lack any evidence of a high-pressure evolution.

A late event of ductile deformation thought to have occurred at around 500 °C (Feybesse et al., 1995; Gardien et al., 1990) juxtaposed the Monts-du-Lyonnais metamorphic complex to the adjacent lower-grade Brévenne Unit (Fig. 2). The last increment of this tectonic phase was marked by the activation of NE-SW trending strike-slip shear zones and the intrusion of several synkinematic granitoids (Feybesse et al., 1995) at 345–335 Ma (Ar-Ar dating on micas and amphibole, Costa et al., 1993). Notably, the Riverie LAC is bounded by such shear zones to the other constituents of the metamorphic complex. Late brittle deformation reworked the original thrust contact between the Monts-du-Lyonnais metamorphic complex and the underlying Pilat unit (representing the parautochtonous terrains).

Little is known about the age of the Monts-du-Lyonnais metamorphic complex protoliths. There are no chronological constraints available on the depositional age of the detrital metasediments. The age of the orthogneiss protolith is estimated to be around 470 Ma (based on zircon Pb evaporation and whole-rock Rb–Sr ages, Dufour, 1982; Feybesse et al., 1995), although Duthou et al. (1984) report a significantly older whole-rock Rb–Sr isochron date of 502 ± 7 Ma. An identical date of 497 ± 8 Ma was obtained for felsic granulites and interpreted as the emplacement age of their igneous protolith (Duthou et al., 1981). Mafic meta-igneous rocks display a variety of geochemical types, ranging from N–MORB to LREE-rich tholeiites and even transitional rocks with alkaline affinities (Briand et al., 1995). They have never been dated, and a Cambro–Ordovician emplacement age has been postulated by analogy with other bimodal meta-igneous associations from the eFMC, notably in the nearby Vivarais area (Chelle-Michou et al., 2017).

In the Riverie belt, felsic rocks described as “amphibole-bearing (felsic) gneisses” and closely associated with amphibolites were first reported by Peterlongo (1958) and later studied by Blanc (1981). The relationships between amphibole-bearing felsic gneisses and amphibolites (the constituent rocks of the belt) are well-exposed in the disused “Les Roches” quarry near Riverie (Fig. 2). Amphibole-bearing felsic gneisses known in other eFMC LACs have been variously regarded as former pyroclastic/epiclastic rocks coeval to the amphibolite protoliths (Pin and Lancelot, 1982), metadiorites (Bodinier et al., 1986) or anatectic magmas formed by partial melting of the amphibolites at Variscan times (Benmammar et al., 2020).

Eight samples of amphibole-bearing felsic gneisses and four samples of amphibolites collected from the pit#2 of the “Les Roches” quarry (GPS coordinates: 45.5964N, 4.5879E, see Supplementary Text) were analyzed for major and trace element compositions by the ALS Global firm (details on the procedures, detection limits and accuracy/reproducibility in Supplementary text, full dataset available as Supplementary Table S1). Recalculation and plotting of the whole-rock geochemical data were performed using the GCDkit plugin for R (Janoušek et al., 2006). Mineral major element compositions were obtained using a Zeiss EVO MA15 Scanning Electron Microscope at the Central Analytical Facility of Stellenbosch University, South Africa. Mineral compositions were determined by EDX (Energy Dispersive X-ray) analysis using an Oxford Instruments® X-Max 20 mm2 detector and the Oxford INCA software. Beam conditions were 20 kV accelerating voltage, 1.5 nA probe current, a working distance of 8.5 mm and a specimen beam current of 20 nA. Analyses were quantified using natural mineral standards and representative mineral compositions are reported in the Supplementary Table S2. Biotite and amphibole structural formulas (including ferric iron contents) were calculated using the spreadsheets of Li et al., (2020a, 2020b). Two samples of amphibole-bearing felsic gneisses (RV-1 and RV-2) were selected for zircon U–Pb–Hf determinations to constrain the emplacement age and isotopic signature of their protoliths. Zircon grains were separated from the crushed rock samples at Saint-Etienne University using conventional techniques (sieving, panning, magnetic and heavy liquids separation followed by handpicking). Selected grains were subsequently cast into epoxy mounts and polished down to a subequatorial grain section. BSE and CL-imaging were performed at the Central Analytical Facility of Stellenbosch University using a Zeiss MERLIN Scanning Electron Microscope. In situ U–Pb dating and zircon Lu–Hf isotope measurements were carried out by LA–ICP–MS at the J.W. Goethe University, Frankfurt-am-Main (Germany). Plotting and calculation of Concordia dates were performed using IsoplotR (Vermeesch, 2018). Information on the analytical methods is presented in the Supplementary Text. The full datasets (standards and samples) are reported in Supplementary Tables S3 to S6.

The Riverie belt occurs as a tectonic boudin bounded by dextral shear zones and syn-kinematic (mylonitic) granites (Fig. 2). In the largest pit of the disused Les Roches quarry, felsic gneisses exhibiting a vertical foliation and wrapping around 1 to 3 m-wide amphibolite bodies (Fig. 3a, b) constitute the main petrographic type. Amphibolites are particularly abundant in the SE part of the pit. Both mafic and felsic rocks are cut across by a network of decimeter-scale discordant or concordant granitic and pegmatitic veins.

The felsic gneisses are primarily composed of plagioclase, quartz, green amphibole with low amounts of brown biotite either in the matrix or rimming amphibole (Fig. 3c, d, e). The foliation is underlined by alternating layers with varying proportions of mafic minerals (Fig. 3c). The felsic rocks show a high-temperature texture with interlobate grain contacts and preservation of high-energy surfaces (Fig. 3d), with amphibole often being interstitial. Inclusions of plagioclase in amphibole and vice versa are locally observed (Fig. 3e). Amphibole is a magnesio-ferri-hornblende with mg# of 0.63 and Fe³⁺/Fe^tot ratios in the range 0.29–0.37, slightly higher than the value (0.27) estimated through wet chemistry by Blanc (1981). Biotite is also magnesian, with mg# of 0.65–0.72, and TiO₂-rich (>3 wt.%). Plagioclase is chemically unzoned and has an oligoclase composition (An_24-26). Accessory minerals are magnetite, apatite, zircon, and sulphides (Fig. 3c, d, e). The edenite–richterite thermometer of Holland and Blundy (1994) and the barometer of Molina et al. (2015) for paired plagioclase–amphibole compositions yielded equilibration temperatures of c. 700 °C for upper to mid-crustal pressures of 2–4 kbar.

The amphibole-bearing felsic gneisses do not deviate from the igneous trend in the FMW plot of Ohta and Arai (2007) which points to an igneous and not volcanosedimentary origin for their protoliths (Fig. 4a). They show SiO₂ contents ranging from 61 to 69 wt.%, reflecting the varying mineral modes at the scale of the quarry (50 m). They are metaluminous to subaluminous with A/CNK (molar Al₂O₃/[CaO + Na₂O + K₂O]) below 1.05, and markedly poor in K₂O (≤1 wt.% and mostly below 0.7 wt.%, Supplementary Table S1). Accordingly, they plot in the field of tonalites in the P–Q diagram of Debon and Le Fort (1983) (Fig. 4b). All samples show similar immobile, incompatible trace elements patterns characterized by overall low concentrations. The amphibole-bearing felsic gneisses are depleted in HREE (by a factor 0.3 to 0.6) and slightly enriched in LREE–Zr (by a factor 2 to 3) and Th (10 times) with respect to N-MORBs (Fig. 4c). Of importance are the pronounced Nb and Ti negative anomalies (with Nb/Nb*: 0.2–0.3; Ti/Ti*: 0.3–0.6), along with a weakly positive Zr anomaly (Zr/Zr*: 1.2–2.1). Th/Nb ratios are high, clustering around 0.6–0.7 (Fig. 4e). The concentrations of LILE such as Ba (100–400 ppm), Sr (200–300 ppm) and Rb (8–20 ppm) are lower than those of the Bulk Continental Crust (Rudnick and Gao, 2003). Chondrite-normalized REE patterns (Boynton, 1984) are moderately fractionated (La_N/Yb_N = 2.9–7.1) and show a slight depletion in mid-HREE (Dy to Tm, Fig. 4d). A weak Eu negative anomaly is observed (Eu/Eu* = 0.71–0.95). Vanadium contents are low (78–151 ppm) with V/Ti ratios close to 20 (Fig. 4f). Most trace element compositions (except Zr) show a rough decrease with the SiO₂ content.

The mafic rocks have a nematoblastic texture and are dominantly composed of an equigranular assemblage of euhedral to subhedral amphibole and plagioclase (Fig. 3f) with apatite and titanite as the main accessory minerals. These amphibolites show mildly alkaline basic compositions (SiO₂: 46.6–51.9 wt.%; Na₂O+K₂O: 4.2–5.1 wt.%, Fig. 4b). Three samples show primitive compositions (i.e. close to that of primary mantle-derived melts) as indicated by their elevated Mg# (62–63), MgO (7.5–9.0 wt.%) and Cr (270–320 ppm) contents and the lack of Eu anomaly. They are slightly enriched in Th and LREE with respect to N-MORBs, and three samples display a pronounced negative Nb anomaly (with Th/Nb>0.2, Fig. 4c, e). REE patterns are nearly flat (La_N/Yb_N = 1.5–2.0, Fig. 4d). Vanadium contents are moderately elevated (227–362 ppm) with V/Ti ratios between 25 and 40 (Fig. 4f).

Several observations suggest that the igneous protolith of the amphibole-bearing felsic gneisses was plutonic and coeval to or younger than that of the amphibolites: (i) the contacts between the two rock types are often intricate, with preserved evidence for intrusive relationships of the former in the latter (see white triangles, Fig. 3a); (ii) pluri-centimetric amphibole-rich clusters in the felsic rocks resemble assimilated mafic material (black triangle, Fig. 3a); (iii) the foliation of the gneisses fades in strain shadows at the edges of the mafic bodies so that the felsic rock appears isotropic (white triangle, Fig. 3b).

Zircon grains extracted from amphibole-bearing felsic gneisses are euhedral to subhedral, ranging in length between 70 and 200 μm, and often show width-to-lengh ratios higher than 1:2 with well-developed pyramidal tips (Fig. 5a). A few grains are stubbier with lower aspect ratios down to 1:1. CL-images reveal concentric oscillatory zoning, locally associated with sector zoning (see for instance zircon 1 of sample RV-2, Fig. 5a). The zoning pattern may differ from the inner to outer part of few grains, due to varying relative growth rate of crystal faces upon crystallization (white triangles on Fig. 5a). However, there is no marked textural break such as a change in CL intensity or clear evidence for resorption. Only very thin (<10 μm-large) CL-bright overgrowths truncate the zoning pattern of some grains. Local recrystallization is also evidenced (black triangles on Fig. 5a).

In gneiss sample RV–1, 40 measurements were performed on 32 grains, all within zones displaying clear oscillatory zoning (Fig. 5a). Seven spots showed moderate common Pb contents (up to 3%) and were markedly discordant. These spots will not be discussed further. A total of 16 concordant spots are statistically equivalent, and a Concordia date of 545.9 ± 1.9 Ma (± 6.4 Ma after propagation of systematic uncertainties) can be calculated from them (MSWD of concordance and equivalence = 1.6). Besides, 6 analyses yielded similar ²⁰⁶Pb/²³⁸U dates but were discordant, presumably owing to small amounts of common Pb. Eleven analyses showed younger, slightly discordant to concordant ²⁰⁶Pb/²³⁸U dates ranging between 536 ± 10 and 509 ± 9 Ma.

In gneiss sample RV–2, 33 analyses were performed on 25 grains. A Concordia date of 543.9 ± 1.5 Ma (±6.3 after propagation of systematic uncertainties) can be calculated from 19 statistically equivalent concordant analyses (with a MSWD of concordance and equivalence equal to 1.4, Fig. 5b). Spot a58 yielded an identical yet discordant ²⁰⁶Pb/²³⁸U date owing to the presence of common Pb. Nine measurements devoid of common Pb yielded concordant to moderately discordant ²⁰⁶Pb/²³⁸U dates ranging from 536 ± 6 down to 520 ± 6 Ma. Two analyses (a50 and a60) characterized by minor common Pb contents (0.3–0.8%) showed discordant and younger ²⁰⁶Pb/²³⁸U dates of 484 ± 5 and 438 ± 7 Ma respectively. Two measurements performed on a single zircon grain yielded a concordant ²⁰⁶Pb/²³⁸U date of 361 ± 4 Ma and a discordant ²⁰⁶Pb/²³⁸U date of 354 ± 4 Ma (estimated common Pb content: 0.4%). Both also featured markedly lower Th/U ratios (0.02 and 0.04, respectively) than grains yielding ²⁰⁶Pb/²³⁸U dates of c. 545 Ma (always >0.15).

Twenty-eight measurements were performed on grain domains analogous to those that previously yielded concordant ²⁰⁶Pb/²³⁸U dates (at ca. 545 Ma) and three on undated grains texturally identical to the main zircon population. Initial Hf isotope compositions were calculated using the intrusion age determined for each sample. The ¹⁷⁶Hf/¹⁷⁷Hf(t) ratios of RV-1 zircons were all identical within uncertainty, ranging from 0.282709 ± 42 to 0.282752 ± 35 (at 2 s), corresponding to εHf(t) of +9.5 to +11.0 with an average value of +10.5 ± 0.9 (2 S.D. − standard deviation; n = 14; Fig. 5c). Zircons from sample RV-2 showed a similar range of ¹⁷⁶Hf/¹⁷⁷Hf(t) ratios, from 0.282717 ± 33 to 0.282787 ± 44 (2 S.E.) corresponding to εHf (t) of +9.7 to +12.2 and yielding an identical average εHf (t) at +10.6 ± 1.2 (2 S.D.; n = 18).

Field relationships, petrographic, and geochemical data indicate that the amphibole-bearing felsic gneisses represent former tonalites (see Sect. 4). Zircon grains show euhedral shapes and oscillatory zoning, as typically observed in zircon grown from melts (Corfu et al., 2003). Hence, the Concordia dates of 545.9 ± 1.9 (6.4) and 543.9 ± 1.5 (6.3) Ma are interpreted as late Ediacaran crystallization ages of the (meta)tonalites RV-1 and RV-2, respectively. The other, discordant, zircon U-Pb data for both samples can be explained by a combination of initial Pb incorporation and/or Pb loss from zircons that originally belonged to the dominant population of concordant, c. 545 Ma-old grains. This interpretation is consistent with the fact that all zircons show identical ¹⁷⁶Hf/¹⁷⁷Hf ratios within uncertainties regardless the apparent ²⁰⁶Pb/²³⁸U date, as disturbance of the U–Pb system does not affect Hf isotopes (Gerdes and Zeh, 2009).

Analyses a65 and a66 were both performed close to the rims of the same zircon grain and showed younger ²³⁸U/²⁰⁶Pb dates of c. 360 Ma, concordant for a65. CL images do not provide clear evidence for the existence of textural discontinuities associated with new zircon growth or extensive zircon recrystallization. Yet, both analyses show very low Th/U ratios, different from the main c. 545 Ma population. This suggests that upon laser ablation both analyses may have (partly or wholly) sampled narrow CL-bright overgrowths or recrystallized domains visible on a few grains (Fig. 5a). Therefore, it is unclear whether the c. 360 Ma dates correspond to the age of a Variscan metamorphic overprint, which would be in agreement with what is observed in the Upper Gneiss Unit of the eFMC (Chelle-Michou et al., 2017), or represent geologically meaningless, mixed ages.

At the scale of the quarry outcrop (approximately 50 m), the (meta)tonalites exhibit a range of major element compositions yet very similar mineral assemblages. Since they derive from a plutonic protolith, these characteristics may reflect different proportions between igneous minerals, resulting from variable extents of crystal accumulation versus crystal/melt separation during solidification of a mush at the emplacement site (e.g.Barnes et al., 2016; Cornet et al., 2022; Laurent et al., 2020; Lee and Morton, 2015). By considering a crystallizing low-K dacitic melt with a chemical composition similar to the average of the (meta)tonalites, the samples can be modeled as a combination of (Fig. 6): (i) residual melts resulting from the separation of 5–20 wt.% plagioclase An₂₅ + hornblende ± magnetite from the initial melt (for samples LY25, RV-1, RV-2); (ii) cumulates, corresponding to the initial melt plus 5–25 wt.% accumulated plagioclase An₂₅ ± hornblende ± magnetite (for samples LY23, −24, 26, −27, −28). Trace element data are consistent with this interpretation given the relatively gradual decrease in most trace elements with SiO₂ contents.

In general, low-K dacitic melts are known to derive from mantle melting followed by fractionation of a primary H₂O-rich basaltic melt or from dehydration melting of amphibolites (Beard, 1995; Stern et al., 1996). Zircon Hf isotope data indicate that the dacitic melt from which the Riverie felsic rocks originated had a composition overlapping with that of the Depleted Mantle (DM) reservoir at c. 545 Ma, ruling out an origin by melting of an old mafic (amphibolitic) crust (Fig. 5c). Therefore, we favor the formation of the dacitic melt through crystal fractionation from a primary mantle-derived melt, consistent with its weak depletion in mid-HREE (Dy to Tm) and the Eu negative anomaly, which are indicative of amphibole and plagioclase fractionation (Bédard, 2006; Nandedkar et al., 2016). The pronounced Nb–Ti negative anomaly (Fig. 4c) and the radiogenic Hf isotope composition (εHf_(545Ma) of c. +11) of the low-K dacitic melt collectively indicate that its depleted mantle source had been metasomatized by fluids derived from an oceanic slab (Fig. 4e; Pearce and Peate, 1995). The presence of magnetite, the elevated amphibole Fe³⁺/Fe_tot as well as the low bulk-rock V/Ti ratios displayed by the (meta)tonalites are consistent with a parental melt having formed under relatively oxidizing conditions (Fig. 4f), which aligns with a metasomatized mantle source.

The country-rocks of the tonalitic intrusion corresponded to gabbros or basalts (now amphibolites). Those have near-primary mantle melt compositions which entails that their negative Nb anomalies can be regarded as pristine and not related to crustal contamination (Fig. 4e). Therefore, the igneous protolith of the amphibolites originated from a mantle having incorporated a slab-derived component, as for the parental melt of the (meta)tonalites, meaning that both mafic and felsic rocks were possibly generated from the same metasomatized mantle source.

As demonstrated in the previous section, the felsic rocks from the Riverie belt can be traced back to a mantle metasomatized by slab-derived fluids shortly prior to melting, namely in the Ediacaran period. At that time, the terrains currently exposed in the eFMC were situated along the northern Gondwana margin, which was significantly influenced by the Avalonian–Cadomian accretionary orogeny (Andonaegui et al., 2016; Collett et al., 2020; Garfunkel, 2015; Linnemann et al., 2014, 2008; Soejono et al., 2017). Therefore, it is likely that the Riverie magmatism tapped into a mantle imprinted by Cadomian subduction. In fact, the Riverie rocks share isotopic similarities with the late Cryogenian–early Ediacaran juvenile arc rocks of the Cadomian type-area in northern France (Samson et al., 2003; Fig. 5c).

According to full-plate tectonic models, the southwards subduction of the Iapetus/Mirovoi oceans was still ongoing at the late Ediacaran/early Cambrian transition (Fig. 7a, Merdith et al., 2021). Consistently, evidence for the existence of 0.55 Ga juvenile arcs along the northern Gondwana margin is provided by the detrital zircon record of the parautochtonous terrains in the eFMC (Couzinié et al., 2019), the Corsica basement (Avigad et al., 2018), the Upper Allochton of the Iberian Massif (Albert et al., 2015), and the Polish Góry Sowie Massif (Tabaud et al., 2021). Additionally, in the Ossa-Morena Zone of the Iberian Massif, juvenile arc-derived magmatism is represented by andesites–dacites interbedded within the late Ediacaran/early Cambrian Malcocinado/San Jerónimo formations and diorites–tonalites from the c. 555 Ma-old Mérida–Montoro igneous complex (Bandres et al., 2002; Pin et al., 2002). Both suites exhibit positive εNd(t) values along with trace element characteristics that resemble the Riverie (meta)tonalites.

In the Maures massif, the Bormes orthogneiss was recently regarded as a remnant of an Ediacaran (590–556 Ma) arc by Tabaud et al. (2023). However, the presence in their samples of 505–460 Ma-old zircon rims together with the major and trace element signature of the peraluminous high-K host rocks rather suggest that they represent Cambrian–Ordovician crust-derived granitoids related to the rifting event that led to the opening of the Variscan oceans (e.g.Couzinié et al., 2022). Arc magmatism in the Maures massif is more likely represented by the Arcs–Gassin leptynites, which have εNd(t) values ranging from +1 to +2 and yielded a protolith emplacement age of 548 +15/-7 Ma (U–Pb on zircon; Innocent et al., 2003). As this age was obtained by isotope-dilution thermal ionization mass spectrometry on multi-grain fractions, it deserves to be confirmed using in situ techniques (Paquette et al., 2017).

The Riverie rocks represent modest volumes and their V/Ti systematics suggest a distal position with respect to the Cadomian arc (Fig. 4f). Therefore, we infer that this juvenile magmatism would have occurred within the back-arc domain of the Cadomian orogen and resulted from the convective thinning of the lithospheric mantle (Fig. 7b).

The Riverie mafic–felsic association represents the first unambiguous record of late Ediacaran magmatism in the Upper Gneiss Unit of the eFMC and is distinctly different from coeval magmatic rocks in the parautochtonous terrains (Fig. 4). In the latter, the c. 550–530 Ma period is marked by voluminous peraluminous felsic magmatism, which corresponds today to the Velay Orthogneiss Formation (Couzinié et al., 2017), the Montredon-Labessonié orthogneiss (Couzinié and Laurent, 2021), the Plaisance–Cammazes orthogneiss (Guérangé-Lozes et al., 2013) and the Rivernous and Sériès dacites–rhyolites in the Montagne Noire area (Lescuyer and Cocherie, 1992; Padel et al., 2017). Collectively, the geochemical (major/trace element and isotopic) signatures (Fig. 4) and zircon inheritance patterns exhibited by these rocks indicate significant crust reworking through the melting of Neoproterozoic sedimentary sequences (Chelle-Michou et al., 2017; Couzinié et al., 2017). The absence of evidence for late Ediacaran compressional deformation (Álvaro et al., 2014) suggests that this crustal melting event took place in an extensional environment, presumably in the back-arc of the Cadomian orogen (Couzinié and Laurent, 2021). The very contrasted typology of late Ediacaran magmatism between the Upper Gneiss Unit and the parautochtonous terrains indicates that they represent two distinct crust segments which had different paleogeographic positions within the late Ediacaran Cadomian back-arc.

The age of magmatism observed in the Riverie belt is unusual compared to other eFMC LACs. The compilation of LA-ICP-MS zircon U-Pb dates obtained from LACs’ meta-igneous rocks (Fig. 5d) indicates that magmatism mostly occurred between 492 and 472 Ma i.e., the Furongian–Lower Ordovician epochs (Chelle-Michou et al., 2017; de Hoÿm de Marien et al., 2023; Lotout et al., 2020, 2018; Paquette et al., 2017; Whitney et al., 2020). However, the rocks from the Riverie belt are at least 50 million years older, dating back to the late Ediacaran period. Furthermore, the meta-igneous felsic rocks of the LACs are generally potassic and display time-integrated unradiogenic Nd-Hf isotope compositions (εNd_(t) < −3; εHf_(t) < −1), which is interpreted as reflecting an origin through melting of the local continental crust (Chelle-Michou et al., 2017; Pin and Marini, 1993). In contrast, the felsic rocks in the Riverie belt are sodic, have radiogenic Hf isotope compositions, and ultimately originated from a juvenile source. The pressure and temperature of mineral equilibration estimated for the Riverie felsic rocks are close to the tonalite solidus (Schmidt and Thompson, 1996) at relatively low pressures (2–4 kbar), possibly reflecting the crystallization conditions of the parental low-K dacitic melt. These rocks do not record evidence for a high-pressure metamorphic evolution, and no eclogite has ever been recovered from the Riverie belt (Blanc, 1981; Feybesse et al., 1995). Collectively, petrographic, geochemical (including isotopic) and geochronological data obtained on the Riverie belt show that the igneous protoliths of this LAC did not form during the rifting of the northern Gondwana margin that resulted in the opening of the Mid-Variscan ocean and were presumably not subducted during the Devonian. These results highlight that all LACs from the Monts-du-Lyonnais, and more generally from the eFMC, are not identical in terms of age, petrogenesis and metamorphic evolution, arguing against their grouping as a single litho-tectonic unit.

The amphibole-bearing felsic gneisses from the Riverie LAC in the Monts-du-Lyonnais metamorphic complex represent a former tonalitic intrusion of late Ediacaran age (c. 545 Ma). The parental melt underwent variable amounts (5–25 wt.%) of crystal accumulation versus crystal/melt separation at the emplacement site and formed through fractionation of a mafic melt that originated from a mantle source metasomatized by slab-derived fluids. The (meta)tonalites are intimately associated with mafic (meta-)igneous rocks, which were likely derived from the same source. The Riverie mafic–felsic association is unrelated to the Furongian (late Cambrian)–Lower Ordovician rifting that resulted in the opening of the Mid-Variscan ocean and instead represents a newly identified remnant of Cadomian arc/back-arc magmatism along the northern Gondwana margin. Overall, our results challenge the assumption that all mafic–felsic associations (LACs) from the high-grade domains of the eFMC can be grouped together as a single litho-tectonic unit (“the” LAC). A detailed reexamination of undated LACs in the eFMC (e.g. the Artense, Sioule, Truyère, Marvejols massifs) using modern techniques is essential for clarifying the diversity of these associations and refining their correlations.

C. Guilbaud made the thin sections. G. Stevens and A. Gerdes granted access to the Stellenbosch CAF and Frankfurt LA–ICP–MS laboratory, respectively. A. Laurie, M. R. Franzenburg and L. Marko assisted during the SEM and LA–ICP–MS sessions. O. Vanderhaeghe provided a template for Figure 1. L.-S. Doucet drew Figure 7a. H. Bertrand introduced the quarry to SC in 2011. Thorough reviews were provided by F. Roger and an anonymous reviewer. The manuscript was handled by V. Bosse and L. Jolivet. We extend our gratitude to all of them.

This work is dedicated to the memory of J.-L. Paquette, our collaborator for a decade, for all his input on topics related to the eFMC, geochronology and granitoids.