Deciphering the nature and age of the protoliths and peak P−T conditions in retrogressed mafic eclogites from the Maures-Tannneron Massif (SE France) and implications for the southern European Variscides

Since the first definition of Haüy (1822), mafic eclogites have long held specific interest for geologists owing to their singular mineral composition and geodynamic significance in mountain belts. Starting with the seminal investigations of Bearth (1959), Miyashiro (1961), Coleman et al. (1965) and Ernst (1971), a consensus has been built around the interpretation of both low-temperature eclogites (group C eclogites of Coleman et al., 1965) and high-temperature, gneiss hosted eclogites (group B eclogites of Coleman et al., 1965) as the metamorphic equivalent of oceanic crust transformed at mantle depths in subduction zones. These pioneering works have been expanded in the last three decades by a significant number of petrologic, chronologic and thermo-mechanical investigations that have shown that in continental collision zones, these metabasites are key targets to understand the early subduction dynamics involved in the formation of mountain belts (see for example Spalla et al., 1996; Gerya and Stöckhert, 2005; Hacker, 2006; Štípská et al., 2006, 2016; Ernst and Liou, 2008; Brown, 2010; Massonne et al., 2018; Gilotti, 2013; Hacker et al., 2013; Collett et al., 2017, 2018; Zhao et al., 2017; Lotout et al., 2020; Roda et al., 2020; Regorda et al., 2021). However in large and hot orogens, such as the European Variscan belt, eclogites can be severely transformed and re-equilibrated under granulite and/or amphibolite-facies conditions, making it challenging to decipher natural and protolith ages, as well as eclogite peak P−T conditions (Dufour et al., 1985; Piboule and Briand, 1985; O’Brien, 1997, 2000; Medaris et al., 1995; Klapova et al., 1998; Faryad et al., 2010; Lardeaux, 2014; Štípská et al., 2014; Cruciani et al., 2015; Scodina et al., 2021). This is particularly the case for the eclogites from the Maures-Tanneron Massif, which is located in southeast France, for which available geochemical, petrologic are equivocal while modern geochronologic constraints are lacking (see Schneider et al., 2014; Oliot et al., 2015). In particular, the prograde/HP stages are poorly constrained, while the decompressional evolution is much better known. Such data are necessary to address the precollision subduction, which is the most obscured Variscan tectonic event, particularly at the southern boundary of this orogenic system that is severely reworked by the Alpine orogeny and opening of the western Mediterranean oceanic basin (Edel et al., 2014, 2018; Spalla et al., 2014; Gosso et al., 2019). Consequently, the goals of this study are the following:

• trace the origin and the tectonic setting leading to the formation of the eclogite protoliths using their geochemical characteristics, including elements that are least mobile during alteration and metamorphism;

• maximize geochronological constraints on the age of the magmatic protoliths and the subsequent metamorphic reequilibrations, using zircon geochronology (LA‒ICP‒ MS/U‒Pb);

• introduce new robust quantitative data on the high-pressure metamorphism, by using trace element contents in specific trapped mineral inclusions in garnets (Ti in quartz and Zr in rutile) and comparing the results with thermodynamic modelling of equilibrium phase assemblages.

The Variscan orogen is an 8000 km-long belt formed as a consequence of Palaeozoic subduction and collision events (Franke, 1989, 1995; Matte, 2001; Faure et al., 2009; Martínez Catalán et al., 2009; Schulmann et al., 2014). Starting with Suess (1926), Kossmat (1927) and Demay (1931) the final orogenic architecture of the European Variscan belt was interpreted as the result of a Carboniferous continent‒ continent collision. Successive episodes of rifting, convergence and collision of several continental microblocks and/or island arcs were involved in the convergence before the final collision of Laurussia and Gondwana (Nance and Murphy, 1994; Scotese et al., 1999; Franke, 2000; Torsvik and Cocks, 2004; Faure et al., 2005; Lardeaux et al., 2014). Finally, late strike-slip faulting driving block rotations led to the actual configuration (Edel et al., 2018).

In Europe, the belt shows an internal orogenic zonation (Kossmat, 1927), which underpins regional-scale correlations and tectonic interpretations (Ballèvre et al., 2009, 2014; Schulmann et al., 2009; Lardeaux et al., 2014; Franke et al., 2017; Martínez Catalán et al., 2020). If there is a general consensus on the existence of a large oceanic domain north of the peri-Gondwanan terranes (i.e., Rheic Ocean), which subducted beneath Laurussia and peri-Gondwana during Devonian times (McKerrow and Cocks, 1995; Cocks and Torsvik, 2002; Stampfli and Borel, 2002; Nance et al., 2010), the existence of one or several successive closures of oceanic and/or back-arc basins during plate convergence is still a matter of debate (Bodinier et al., 1988; Pin, 1990; Matte, 1991; Linnemann et al., 2007; Schulmann et al., 2009, 2014; Lardeaux et al., 2014; Franke et al., 2017; Regorda et al., 2020).

In this general framework, the Maures-Tanneron Massif (MTM), located in southeast France (Fig. 1A), is a segment of the southern part of the Variscan Realm whose geological affinities with Sardinia and Corsica have been recognized for several decades (Carmignani et al., 1992; Crévola and Pupin, 1994; Lardeaux et al., 1994; Carosi and Oggiano, 2002; Bellot, 2005; Elter and Pandeli, 2005; Corsini and Rolland, 2009; Rossi et al., 2009; Schneider et al., 2014; Cruciani et al., 2015). These assumptions are now widely confirmed by high-resolution palaeomagnetic investigations that definitely established that the Maures-Tanneron Massif belongs to the singular microplate called “Maures-Estérel-Corsica-Sardinia” (MECS, Edel et al., 2014, 2015, 2018).

At the regional scale, the MTM can be subdivided into two main N‒S oriented metamorphic domains, which were folded and moderately deformed during north‒south shortening related to the so-called “Pyrenean-Provence” alpine phase (Fig. 1B; Crévola and Pupin, 1994; Bellot, 2005; Corsini and Rolland, 2009; Schneider et al., 2014). The main Variscan metamorphic domains are as follows:

• a very low to medium-grade western domain, mainly consisting of Chlorite‒Chloritoid-bearing phyllites, quartzites and amphibolite-facies micaschists and orthogneiss;

• a high-grade eastern domain composed of migmatitic para- and orthogneiss, with lenses of metabasic rocks with relics of eclogite-facies metamorphism and serpentinized mantle-derived peridotites.

The two domains are separated by the regional-scale La Garde-Freinet/Cavalaire shear zone (Fig. 1B), which is a west-dipping ductile shear zone, considered as the major tectonic suture of the MTM and reactivated as a transpressive shear zone at the end of the orogenic history (Bellot et al., 2002, 2003; Corsini and Rolland, 2009; Rolland et al., 2009; Gerbault et al., 2018; Simonetti et al., 2020).

The polyphased tectono-metamorphic evolution of the MTM can be summarized as follows (Schneider et al., 2014; Oliot et al., 2015): an early oceanic subduction event (unknown age) revealed by relics of mafic eclogites found only in the eastern domain. This subduction was followed, during the Devonian–Mississippian times (i.e., 360–330 Ma), by continental collision, leading to NW-directed (in present day coordinates) nappe tectonics (D₁) and Barrovian-type metamorphism (M₁, i.e., 5–7.6 kbar and 430–610 °C in the external domain and up to 9–11 kbar and 600–650 °C in the internal domain; Buscail, 2000; Bellot, 2005; Bellot et al., 2010), then by back thrusting of the nappes towards the SE (D₂). The D₂ phase developed under medium-pressure and medium- to high-temperature metamorphic conditions mainly in the eastern domain (M₂, i.e., 5–6 kbar and 560–618 °C in the external domain and 7–8 kbar for 750–850 °C in the internal domain, Buscail, 2000; Bellot et al., 2010; Schneider et al., 2014). During the Pennsylvanian (i.e., 330–300 Ma), the latter was affected by highly partitioned transpression (D₃ phase), with an ca. E‒W shortening, producing plurikilometric concentric folds associated with an ca. N‒S strike-slip shear zone, as the Grimaud–Joyeuse and Lamoure faults. The D₃ transpressional phase was associated with the exhumation of metamorphic units, low-pressure/high-temperature metamorphism (M₃, i.e., 4–5 kbar for 600–650 °C, evolving to 2–5 kbar and 300–450 °C; Bellot et al., 2003; Corsini and Rolland, 2009; Schneider et al., 2014) and formation of migmatites, as well as the emplacement of granitoids at approximately 325–300 Ma (Morillon et al., 2000). This evolution ends with the opening of the intramountainous Middle to Upper Pennsylvanian (i.e., 305–300 Ma), coal-bearing, continental sedimentary basins sometimes associated with intrusions of microgranite dikes.

One of the most remarkable elements of the geology of the MTM is the occurrence of several leptynite–amphibolite complexes (LAC, Bard and Caruba, 1981; Seyler, 1986; Bouloton et al., 1998; Bellot et al., 2002). The LAC is a distinctive rock association that is well-known in the French Variscan belt (Forestier et al., 1973; Piboule and Briand, 1985; Santallier et al., 1988; Ballèvre et al., 2009; Faure et al., 2009; Lardeaux et al., 2014) and is defined by the association of mafic rocks (eclogitized meta-basalts or meta-gabbros, amphibolitized ophiolites and mafic HP granulites…) and felsic rocks (meta-rhyolites and meta-granites) sometimes with spinel and/or garnet peridotites. In the western part of the MTM (Fig. 1B), this bimodal magmatism, of unknown age, is interpreted as the relic of pre-Variscan, possibly plume-induced, continental rifting (i.e., active rifting, Seyler, 1986; Briand et al., 2002). In this Western domain (“Collobrières unit”, Seyler, 1975; Seyler and Boucarut, 1978), the LAC displays an alkaline affinity, with metavolcanics metamorphosed under upper greenschist to amphibolite-facies conditions, whereas in the central part of the MTM (Les Arcs−La Garde Freinet), it displays tholeiitic to transitional signatures (Seyler, 1986; Bellot et al., 2010). In the easternmost part of the MTM (i.e., east of the Joyeuse-Grimaud strike-slip fault, Fig. 1B), eclogites occur as numerous boudins, sometimes associated with serpentinites and/or metagabbros, within metasediments (Laverne et al., 1997; Bellot et al., 2010; Schneider et al., 2014). In this case, the metabasites display tholeiitic affinities that are typical for an oceanic crust and/or oceanic back-arc basin. This regional-scale geochemical zonation is interpreted as the relic of a progressive transition from a rifted continental margin to oceanic crust (see review in Schneider et al., 2014). The mafic rocks are metamorphosed under eclogite-facies (Figs. 1B and 2) and severely retrogressed under amphibolite-facies conditions (Bard and Caruba, 1981, 1982; Seyler and Crevola, 1982; Caruba, 1983; Crévola and Pupin, 1994). Consequently, if retrograde evolutions of basic rocks are well documented, the pressure peak conditions of the MTM mafic eclogites are still poorly constrained (see Schneider et al., 2014 for a synthesis of the calculated P−T paths with references therein). In this eastern MTM, the age of amphibolite-facies metamorphism is constrained at 348 ± 7 Ma and 358 ± 7 Ma (Rb–Sr whole-rock; Innocent et al., 2003 and U–Th–Pb (EPMA in situ dating) on monazite, Oliot et al., 2015, respectively), with progressive cooling ages ranging from 329.9 ± 2.1 Ma to 302.8 ± 2.3 Ma (⁴⁰Ar–³⁹Ar dating of brown amphibole; Morillon et al., 2000; Corsini et al., 2010).

South of Saint Tropez, at Cap Pinet, migmatitic paragneisses with decametric mafic lenses crop out along the coastline (Fig. 2). The foliation dips very steeply with a NNE‒SSW orientation, bears a subhorizontal stretching lineation, and corresponds to the main regional S2 plane that transposed during the D₃ event. The pronounced tubular shape of the amphibolite bodies (Fig. 3) results from intense stretching. In this area, the finite strain pattern and thus the observed tectonic structures are associated with the D₃ transpression phase, which reorients the structures inherited from the previous D₁ and D₂ phases (Schneider et al., 2014). As gneisses and amphibolites are both affected by the same D₃ deformation they were likely mixed during the early evolution stages of the MTM (see Gerbault et al., 2018 for discussion).

Because we focus on eclogites, all the studied samples were collected in the eastern domain of the MTM. In the field (Fig. 3), most of the metabasites occurr as foliated amphibolites (Fig. 4A and B), sometimes with relicts of garnet that are partly replaced by coronas of plagioclase and amphibole (Fig. 4C and D). In thin sections, these amphibolites display discrete relicts of Ca-clinopyroxene + plagioclase symplectite (Sect. 4.2.1 Petrography of eclogites), a microstructure that is diagnostic of omphacite retrogression (Boland and Van Roermund, 1983; Joanny et al., 1989, 1991). Thus, amphibolites are retrograded amphibolite-facies equivalents of eclogites. However, the best-preserved metabasites are found only as undeformed relicts in the core of metric to decametric boudins within foliated rocks. At the regional scale the best-preserved mafic boudins outcrop in the so-called “Tahiti Beach” (Fig. 3), near Saint-Tropez.

Mineral chemistry was determined by the electron probe microanalyser CAMECA SX 100 at the Laboratoire Magmas et Volcans, Université Clermont-Auvergne, France and at the “Service Commun de Microsonde” of the Montpellier University, France. In both cases, the operating conditions were a 20 kV accelerating voltage, a 10 nA for beam current, a 1 µm for beam size and a counting time of 10 seconds for analyses of silicates except for quartz inclusions in garnets for which conditions of a 20 kV accelerating voltage, a 100 nA beam current and a counting time of 120 seconds were used to measure titanium.

Peak eclogitic conditions were calculated only in undeformed and best-preserved garnet-bearing samples (i.e., samples MTME3, MTME4 and MTME5). In the latter, we studied the mineral inclusions trapped within the cores of garnet. These inclusions, which were first observed with a petrographic microscope, were also characterized using scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX) at the “Centre commun de microscopie appliquée” of the Côte d’Azur University using a Tescan Vega 3 XMU SEM system (TESCAN FRANCE, Fuveau, France) equipped with an X-MaxN 50 EDX detector (Oxford Instruments, Abingdon, Great Britain).

In our samples we first used the Zr-in-rutile thermometer, as the solubility of trace elements within metastable rutile inclusions can be useful indicators of frozen P−T conditions (Watson et al., 2006; Tomkins et al., 2007). For this purpose, in thin sections, rutile was identified in reflected and transmitted light using a petrographic microscope to select the larger grains showing no transformation into ilmenite. Rutile inclusions were also backscattered electron (BSE) imaged to select grains without any trace of replacement by ilmenite using scanning electron microscopy (SEM) with an accelerating voltage of 20 kV. The selected well-preserved rutiles (i.e., without ilmenite corona or exsolution needles) were analysed using a CAMECA SX100 electron microprobe with operating conditions of a 20 kV accelerating voltage, a 100 nA current and a counting time of 90 seconds. A 1 µm spot beam was used on grains up to 0.2–0.4 mm in length. Furthermore, we considered the possibility of obtaining direct and robust calculations of P−T conditions of inclusion crystallization using the Zr-in-rutile thermometer together with the Ti-in-quartz thermobarometer (Thomas et al., 2010) on contemporary rutile and quartz inclusions preserved in garnet cores. We also performed thermodynamic modelling using the free energy minimization program THERIAK/DOMINO (De Capitani and Petrakakis, 2010, version 2015) on the least retromorphosed sample (MTME3), even if in retrograded eclogites the effective rock composition with which the garnet was in equilibrium remains difficult to define (see Lanari and Engi, 2017, for discussion). Thus, even if questionable, the modelled composition was determined using the whole-rock analysis. We used the internally consistent thermodynamic “tcdb55c2d” database (Holland and Powell, 2004) and the following mineral a–x relations: plagioclase (Holland and Powell, 2003), garnet (White et al., 2007), ilmenite (White et al., 2000), amphibole (Diener et al., 2007), clinopyroxene (Green et al., 2007), orthopyroxene (White et al., 2002), biotite, (White et al., 2007) and white mica (Coggon and Holland, 2002).

For LA‒ICP‒MS/U‒Pb zircon geochronology, we also selected the best-preserved MTME3 sample. Zircon grains were separated using standard heavy liquid and magnetic techniques. Prior to analysis back-scattered electron (BSE) and cathodoluminescence (CL) images of zircons were obtained at the Czech Geological Survey. These images were used to choose spot locations for LA‒ICP‒MS analyses. LA‒ICP‒ MS/U‒Pb analyses on zircons were conducted at the Czech Geological Survey, Prague and at the GeOHeLiS Platform, University of Rennes, France, and the procedure is described in detail in the Supplementary material (Analytical Methods). Only zircon analyses with 98–102% concordance and not affected by Pb loss or inheritance (Gehrels, 2012) were used. Probability density plots of analytical data were generated using “Density Plotter” (Vermeesch, 2012) and Wetherill diagrams using Isoplot 4.15 (Ludwig, 2011).

Whole-rock major and trace element analyses were obtained by using inductively coupled plasma atomic emission spectrometry (ICP‒AES) and inductively coupled plasma‒ mass spectrometry (ICP‒MS) respectively, at the Geochemical and Petrographic Research Center in Nancy (SARM laboratory, CNRS-CRPG) following the procedure proposed by Carignan et al. (2001). To evaluate the effects of amphibolite-facies retrogression on the eclogite chemical composition, both undeformed best-preserved samples and retrogressed samples were analysed. However, the geochemical characterization of protoliths took only the best-preserved samples into account.

Analytical results are presented in the Supplementary material (Table S1).

In the TAS classification diagram (Le Bas et al., 1986), all samples plot within the basaltic field (Fig. 5A) while in the AFM compositional scheme (Irvine and Baragar, 1971), they plot in the field of tholeiitic series, except for one of the severely retrogressed samples (Fig. 5B). However, because the MTM eclogites experienced high-temperature re-equilibration and are embodied within migmatitic paragneiss, significant mobility of major elements and crustal contamination must be considered. Indeed, in comparison with average N-MORB or E-MORB major element compositions (McKenzie and O’Nions, 1991), the MTM samples show, for nearly constant SiO₂ contents (48.5 to 51.6 wt%), sharp depletions in MgO and CaO or clear enrichment in mobile elements (i.e., K₂O, Na₂O or TiO₂), which are notably pronounced for the most retrogressed eclogites. Thus, classification diagrams based on relatively immobile trace elements are preferred, particularly the Zr/Ti versus Nb/Y diagram of Pearce (1996). In the latter (Fig. 5C), a distinction between preserved and retrogressed samples can be observed. When preserved, eclogites plot mainly in the tholeiitic basalt field, while retrogressed samples exhibit significant dispersion, even regarding reputedly immobile elements. This type of evolution has been noted in other examples of retrogressed eclogites and interpreted as tracers of chemical contamination, although the origin of this contamination is still debated (Thompson et al., 1983; Štípská et al., 2014; Cruciani et al., 2015; Jouffray et al., 2020).

In various discriminant diagrams, well-preserved eclogites show typical MORB compositions, whereas the most retrogressed samples trend towards different geotectonic fields (Fig. 5D and E). Focusing on preserved metabasites, in the chondrite-normalized Th versus Nb of Saccani (2015), as well as in the Hf/3-Th-Ta compositional scheme proposed by Wood (1980), the MTM eclogites show E-MORB affinity (Fig. 5F and G). In the Th/Yb versus Nb/Yb diagram elaborated by Pearce (2008) the preserved samples still plot along a MORB-axis, close to E-MORB affinity (Fig. 5H), while a significant shift is observed for the most retrogressed eclogites, indicating possible crustal contamination or a contaminated basaltic source (i.e., contamination of liquids produced during subduction).

Moreover, N- and E-MORBs have uniform Zr/Hf and Nb/Ta ratios of approximately 36 and 17, respectively (Sun and McDonough, 1989). In the most preserved MTM eclogites, the mean values of Nb/La, Zr/Hf and Nb/Ta ratios are 0.71 ± 0.1, 37.8 ± 2.2 and 12.49 ± 0.6, respectively.

Furthermore, the Ce/Pb ratios of approximately 25 ± 5 are characteristic of average MORB compositions (Hofmann et al., 1986), while the Nb/U ratios of 47 ± 11 and Pb/Nd ratios of 0.04 to 0.07 are typical for MORBs (Hofmann, 2003). In our preserved eclogites mean values of 10.3 ± 3.1 for Ce/Pb, 12 ± 2.7 for Nb/U and 0.14 ± 0.05 for Pb/Nd ratios testify to crustal contamination (see Rudnick and Fountain, 1995 with references therein).

Average N-MORB and E-MORB compositions provide reference ΣREE values of 39.11 and 49.1 ppm, respectively (Sun and McDonough, 1989; McKenzie and O’Nions, 1991). The most preserved MTM eclogites show a value of 59 ± 8 ppm, and thus a coefficient of enrichment on the order of 1.6. By comparison, in the severely retrogressed samples this coefficient of enrichment reaches a value of 4.

In the best-preserved MTM eclogites, the Eu/Eu* ratio (commonly corresponding to Eu_N/(Sm_N x Gd_N)^1/2, Costa et al., 2021), which yields relatively homogeneous values (0.95 to 0.98), is consistent with the typical reference value proposed for basalts (0.95 after Philpotts and Schnetlzer, 1968). Most scattered values (0.86 to 1.06) are calculated in the most retrogressed samples. This former indicates homogeneous protolith basaltic chemistry and the latter corroborates the link between the modification of sample chemistry and the degree of retrogression.

Patterns of the primitive mantle-normalized rare earth elements (REEs, Fig. 6A) show a global negative slope (La_N/Yb_N = 2.5–3.5) and enrichment in LREEs with respect to HREEs (La_N/Sm_N = 1.94–2.23 and Sm_N/Yb_N = 1.46–1.72). The patterns of the most preserved eclogites are consistent with the E-MORB trend.

HFSE patterns normalized to the primitive mantle (Fig. 6B), are characterized by depletions in Pb, Y, Sn and Mo or enrichments in La, U or Ta compatible with the E-MORB trend. However, even the best-preserved samples show depletion in Sr or enrichment, which is slight in Nd, indicating probable changes in the initial chemical composition of the protolith.

The MTM eclogites experienced intense retrogression, and preserved typical eclogitic assemblages were never observed in thin sections. The following description is focused on the petrography of the best-preserved eclogites.

The most preserved eclogites are composed predominantly of fine-grained symplectite of plagioclase and Ca-clinopyroxene and garnet surrounded by coronas of plagioclase and amphibole (Figs. 7 and 8). Omphacite single grains were never observed, even as inclusions within garnet, but their prior occurrence is attested to by the fine-grained symplectite (Boland and Van Roermund, 1983; Joanny et al., 1989, 1991).

Garnet occurs as millimetre (1–2 mm), sometimes centimetre, sized grains, commonly rimmed by coronas of brown amphibole, plagioclase and sometimes ilmenite. The best-preserved garnet grains display numerous inclusions of quartz, amphibole and rutile (Figs. 7D and 8). Even when included in garnet, some of the rutile inclusions can be partly transformed into ilmenite. Some garnets are fractured and cracks are filled by amphibole, quartz, epidote and ilmenite. In the MTME3 sample, the high-resolution observation of garnet inclusions using scanning electron microscopy (SEM) coupled with energy-dispersive X-ray spectroscopy (EDX) reveals the presence of preserved white mica (Fig. 8B). This inclusion was analysed using the low vacuum mode of the SEM instrument, and EDX spot analysis revealed the K content in the mica, which is thus a muscovite.

Amphibole represents the most abundant phase and displays four microtextural sites (Fig. 7): (i) brown inclusions within garnet, (ii) in coronas, in association with plagioclase, around garnet, (iii) as replacements of Ca-clinopyroxene in fine-grained symplectite, and (iv) poikiloblasts of brown to greenish grains replacing garnet and/or symplectite.

Zircon is observed as inclusions within garnet but also in symplectite assemblages associated with Ca-clinopyroxene, plagioclase and/or amphibole. Rare biotite is observed in the amphibole + plagioclase-bearing coronas around garnet.

Analytical results are presented in the Supplementary material (Table S2).

Garnet is almandine-rich (46.5–58.6 mol%) with pyrope and grossular contents of 15.6 to 32.7 and 14 to 27.9 mol%, respectively, if we consider all the preserved samples. Garnet compositions plot in the fields of B and C eclogite types (Fig. 9A) in the diagram of Coleman et al. (1965). In the garnets from the MTME3 sample, zoning is very slightly pronounced. In the MTME1 sample, a zonation is depicted with, from core to rim, a decrease in the almandine content associated with an increase in the pyrope content. On the other hand, in the MTME2 and MTME4 samples, garnets display zonation with a decrease in the pyrope content from core to rim. A full compositional profile obtained on a garnet from the MTME1 (Fig. 9B) sample shows that the chemical composition obtained in the garnet core remains relatively constant over a significant distance (ca. 150 µm). The zonation is characterized by an increase in X_Mg (15.5 to 22.5% with X_Mg = Mg/(Mg + Mn + Fe + Ca)) associated with a decrease in X_Mn (4.21 to 0.44% with X_Mn = Mn/(Mg + Mn + Fe + Ca)) from the core to the rim of the mineral, while the variations in X_Fe and X_Ca are much more irregular.

Clinopyroxene in symplectite displays a compositional evolution from omphacite, with jadeite content of approximately 28–30 mol% in coarser lamellae, to mainly diopside in association with plagioclase at the edge of symplectite (Fig. 10).

Plagioclase lamellae (Fig. 10C) associated with omphacite are oligoclase (70–85%mol of albite). When associated with amphibole in coronas around garnet and/or replacing symplectite, they correspond to andesine (63–71%mol of albite).

Amphibole is calcic in the classifications of Leake et al. (1997, 2004) and Hawthorne and Oberti (2012) (Fig. 10D). Amphibole in inclusions within garnet (Amp-I) is tschermakite (#Mg between 0.53 and 0.60). Amphibole in coronas around garnet (Amp-II) and amphibole replacing Ca-clinopyroxene in symplectite (Amp-III) are magnesio-hornblende (#Mg within the range of 0.58 to 0.70). Late poikiloblasts of amphibole (Amp-IV) are actinolite with #Mg value of 0.69.

Based on the previously presented observations, the inferred primary assemblage of the studied eclogites is garnet + omphacite + rutile + amphibole + quartz ± white mica ± zircon.

It is now clearly established that Zr solubility in rutile in equilibrium with quartz and zircon, is strongly temperature dependent, with a moderate P dependence (Degeling et al., 2001; Troitzsch and Ellis, 2004; Zack et al., 2004; Watson et al., 2006; Ferry and Watson, 2007; Tomkins et al., 2007). An important advantage of this method is that Zr is very dilute in rutile. Therefore, the uncertainties related to the nonideality of solid solutions in thermodynamic calculations remain negligible in both phases (Watson et al., 2006). However, the accuracy of the Zr-in-rutile thermometer can be assessed only if (1) rutile is perfectly preserved without any reaction rim with the host mineral (Ewing et al., 2013), (2) quartz and zircon are also present as trapped inclusions in the same host mineral, (3) the Si content in rutile is lower than 200 ppm, and for Si values higher than 200 ppm, the zirconium content could be merely the total of the zirconium in rutile mixed with zirconium derived from nearby phases (see the discussion in Zack et al., 2004), and (4) high values for accelerating voltage, beam current and counting time are implemented (Batanova et al., 2018). Considering all these limitations, we first applied the calibrations of Tomkins et al. (2007), for the Zr-in-rutile thermometer, and of Thomas et al. (2010) for TitaniQ thermo-barometer, in mineral inclusions trapped in well-preserved garnets. In the MTME3 sample, 2 quartz inclusions and 4 inclusions of rutile were analysed, each analysis was duplicated, and each result presented is the average of two analyses. In the MTME5 sample, 3 inclusions of quartz and 3 inclusions of rutile were analysed following the same procedure (the analytical results are presented in the Supplementary material, Table S3).

The values for Ti-in-quartz vary between 5.1 and 6.9 ppm, while for Zr-in-rutile the measured values range between 101.4 and 197.7 ppm. The combination of these data leads to a stability domain of 12–19 kbar for 600–670 °C (Fig. 11).

A P−T pseudosection was calculated with the THERIAK/DOMINO program (De Capitani and Petrakakis, 2010, version 2015) in the SiO₂-Al₂O₃-TiO₂-FeO-MgO-CaO-Na₂O-H₂O system using the whole rock composition of the best-preserved sample (MTME3). Because the MnO content is very low and biotite is clearly a retrogressive phase occurring as interstitial grains or replacing garnets, MnO and K₂O were removed from the modelled composition. As a first investigation, H₂O was considered in excess. In this case, the inferred primary mineral assemblage corresponds to a calculated field of 16.5–22 kbar and 600–690 °C (Fig. 12A). It is bounded by the stability of zoisite at low temperatures and the disappearance of rutile and amphibole at low and high pressures respectively. The combination of the two calculation methods used yields a restricted P−T field of 16.8–18.5 kbar and 630–660 °C for the MTM eclogites (Fig. 12A). However, this H₂O-saturated pseudosection predicts the presence of paragonite in the high-pressure assemblages. This mineral has not been observed, although the inclusion of white mica was identified in garnet.

Because the effect of Fe³⁺ cannot be neglected in modelling mafic rocks, a second pseudosection was calculated. According to Sun and McDonough (1989) and Rebay et al. (2010), a range of values between 9 and 12% Fe₂O₃ of total FeO was considered for basaltic composition. With Fe₂O₃ values higher than 10%, epidote is always present in the metamorphic assemblage of interest, a mineral that has not been observed as trapped inclusions in garnet and thus discards these conditions. On the other hand, for a value of 9% Fe₂O₃ of total FeO, epidote is no longer present in the eclogitic assemblage and the main mineral stability fields identified in the previous models are, at the first order, retained (Fig. 12B). The major change concerns the presence of haematite (1.9–5.2%) and the presence of muscovite instead of paragonite in the metamorphic assemblage is considered the most likely. First, the results of this model are consistent with the P−T conditions calculated in trapped inclusions, and the combination of the two methods allows us to propose a P−T range of 17.2–18.5 kbar for 640–660 °C.

We also tested the effect of water undersaturation on our models. With a value of 9% Fe₂O₃ of total FeO, a model with only 5% H₂O in the system was investigated. In this case, epidote and/or paragonite are always present in the metamorphic assemblage of interest. Moreover, the calculated P−T field for the least different mineral assemblage from that observed is less compatible with the results obtained with the study of trapped inclusions compared to previous models.

Finally, if we accept the postulates of H₂O-saturated conditions and 9% Fe₂O₃ of total FeO for the effective rock composition, the combination of the related pseudosection with the P−T range calculated with trapped quartz and rutile inclusions yields to a P−T estimate restricted to 17.2–18.5 kbar and 640–660 °C for the formation of the MTM eclogites.

The analytical results are presented in the Supplementary material (Table S4).

More than 150 zircon grains, which were all selected from the best-preserved sample, MTME3, were mounted and analysed by laser ablation. Zircons are mostly subrounded (∼50–200 μm) and multifaceted grains (Corfu et al., 2003). Under cathodoluminescence (CL), the majority of zircon grains show CL-grey to dark cores, often overgrown by large CL-grey to dark rims that are structureless (Fig. 13). Only a few grains show CL-dark oscillatory-zoned cores, which are characteristic of igneous zircons (Hoskin and Schaltegger, 2003 with references therein). Limits between internal and external domains are often marked by irregular features (Corfu et al., 2003; Fig. 11C).

The U‒Pb Concordia diagram, kernel density and bar plots of concordant data are represented in Figure 13. The reported errors are all 2σ. The sample displays a pattern of U‒Pb dating characterized by a discontinuous spread of dates from 510 Ma to 310 Ma. Older zircon domains at ca. 710 Ma are extremely rare (Fig. 13B). The few concordant zircon analyses obtained from CL-dark oscillatory-zoned cores define a ²⁰⁶Pb/²³⁸U weighted mean age of 499.5 ± 2.9 Ma (MSWD = 0.88; n = 4). This zircon core group is characterized by homogeneous Th/U ratios (0.3–0.4), which are consistent with a magmatic origin (Rubatto and Gebauer, 2000; Kirkland et al., 2015).

Poorly constrained dates from 450 to 410 Ma are observed in a few zircon domains characterized by low Th/U ratios (Th/U < 0.1), and are thus possibly of a metamorphic origin.

A ²⁰⁶Pb/²³⁸U weighted mean age of 332 ± 2.1 Ma (MSWD = 1.6; n = 15) can be calculated from CL grey cores. This main core population is characterized by a low and restricted range of Th/U ratios (<0.1). Large CL grey to dark rims yield a younger ²⁰⁶Pb/²³⁸U weighted mean age of 321.9 ± 1.5 Ma (MSWD = 0.83; n = 12).

Considering the nature of the protolith of the studied eclogites, major and trace element analyses display a clear affinity with tholeiitic basalts. Our results are clearly in line with previous geochemical studies regarding the eclogitic boudins recognized in the easternmost part of the MTM (Caruba, 1983; Seyler, 1986; Bouloton et al., 1998; Bellot et al., 2010; Schneider et al., 2014). In the study area, i.e., the eastern MTM, both metabasalts and metagabbros are identified and are frequently associated with serpentinites, marbles and siliceous-rich meta-sediments (Caruba, 1983). Moreover, contrary to that is observed in the western part of the MTM (Seyler, 1986; Briand et al., 2002), the metabasites of interest are not associated with metarhyolites (i.e., felsic protoliths possibly derived from continental crust melts). All these petrographic data lead to interpretation of eclogites as relics of oceanic crust rather than witnesses of basaltic dikes that intruded within a thinned continental as envisaged in some of the external crystalline massifs in the Western Alps (Vanardois et al., 2022).

Discriminant diagrams are consistent with the MORB signature, while HFSE and REE patterns seem relatively compatible with an E-MORB affinity. However, ∑REE enrichment relative to N-MORB is of a factor close to 2 for most preserved samples and of 4 for the more retrogressed samples. Significant chemical shifts are observed in various geochemical diagrams in the latter, indicating probable chemical contamination. Consequently, if the MORB affinity should be assumed to be robust, the shift between N-MORB and E-MORB affinity is the result of chemical contamination rather than the witness of a geodynamic setting.

By comparison, the sardinian eclogites preserved within amphibolite-facies gneisses show clear affinity with N-MORBs (Cruciani et al., 2015; Fig. 5H), while eclogite boudins enclosed within migmatites display a clear shift towards E-MORB interpreted as evidence of significant crustal contamination by several authors (Loi, 2003; Cortesogno et al., 2004; Giacomini et al., 2005; Utzeri, 2007; Cruciani et al., 2010). Indeed, during crustal melting, crust-derived fluid influxes within mafic formations have been described (see Pattison, 1991 and Štípská et al., 2014, with references therein). Two working hypotheses must therefore be considered for the MTM eclogites: (i) protoliths are truly E-MORBs (i.e., contaminated basaltic liquids during subduction); thus, supra-subduction zone basalts or (ii) protoliths are N-MORBs, and the derived eclogites are contaminated by the surrounding migmatites during their emplacement. Regardless of the chosen model, the MTM eclogites represent the remnant of oceanic basaltic crust involved in a subduction zone environment.

The zircon cores with typical igneous characteristics show a main U‒Pb age at 499.5 ± 2.9 Ma, which could represent the most likely formation age of the oceanic basaltic crust. Currently, unified full plate reconstruction models are available (i.e., palaeomagnetic databases consistent with coherent plate boundary kinematics, mantle dynamics and geologic features; see Domeier and Torsvik, 2017 for a discussion with references) and offer the most robust framework for palaeogeographic reconstruction in the considered time span. In all the available unified models (Fig. 14) at 548–500 Ma the drift of Avalonia and of the other northern Gondwana microplates has not yet started. Consequently, the volcanic protolith of the MTM eclogites was part of an old oceanic floor located north of the Gondwana active continental margin, called Mirovoi, or the Ran Ocean depending on the authors (Domeier, 2016; Merdith et al., 2021). The opening of the Rheic oceanic domain occurred later during the Ordovician, following the Cambro-Ordovician continental rifting leading to Avalonia drift (Murphy et al., 2006; Linnemann et al., 2007, 2008; Nance et al., 2010, 2012; Torsvik and Cocks, 2011; Franke et al., 2017).

Zircons characterized by low Th/U ratios observed in our sample with Ordovician ages (approximately 450 Ma) are poorly constrained (uncertainty of 10–20 Ma). Because Th/U ratios in this population of zircon rims are generally low (<0.1), they could indicate a metamorphic zircon precipitation (Rubatto, 2002). Moreover, in the studied sample, ²⁰⁶Pb/²³⁸U weighted mean ages recorded in CL grey cores (332 ± 2.1 Ma) or CL fine grey to dark rims (321.9 ± 1.5 Ma) in zircons with Th/U ratios <0.1 attest to the metamorphic evolution during Carboniferous crustal thickening that is well established in the MECS microplate (Rossi et al., 2009; Giacomini et al., 2008; Morillon et al., 2000; Corsini et al., 2010; Li et al., 2014; Faure et al., 2014; Oliot et al., 2015), thus metamorphic zircon precipitation between ca. 350 and ca. 305 Ma could be considered. More specifically, in the internal domain of the MTM, the development of the regional-scale foliation (M1 and/or M2 tectonic events), which clearly postdates eclogite-facies metamorphism, is dated in the range of 360–330 Ma by different methods (Corsini et al., 2010; Oliot et al., 2015). However, considering the spread of U‒Pb data between the two populations, it is also possible that all these analyses do not define a statistically valid age but result from a natural spread of dates, in addition to analytical uncertainties, which is actually reported for metamorphic zircons in Variscan eclogites (e.g.Schmädicke et al., 2018; De Hoÿm de Marien, 2019; Pitra et al., 2022).

Regarding magmatic protolith ages, it seems important to emphasize that in Sardinia, all the protolith ages of the eclogites are Ordovician (470–450 Ma) and thus interpreted as subducted pieces of the Rheic Ocean (Giacomini et al., 2005; Cruciani et al., 2013, 2015). Our results indicate older protolith ages. Thus, the eclogites recognized in the internal metamorphic domain (i.e., the Mid-Variscan Allochthon following Martínez Catalán et al., 2021) of the MECS microplate represent the closure of oceanic domains of different ages. Indeed, at ca. 500 Ma, the MECS microplate, as well as all the other components that amalgamated further during the Variscan orogeny, is located to the north of the Gondwana active margin (see Linnemann et al., 2007, 2008, 2014; Avigad et al., 2012; Margalef et al., 2016; Couzinié et al., 2017; Stephan et al., 2019; Collett et al., 2020; Tabaud et al., 2021, 2022). The location of MTM in the northeast of the active margin of Gondwana was recently demonstrated by the study of inherited zircons identified within the orthogneiss of this massif (Tabaud et al., 2022). At that time, there was only one oceanic domain, located to the north of the considered continental ribbons (or microplates), which was being closed. Moreover, as previously underlined, there is also a broad consensus that Avalonia drift led to the opening of the Rheic oceanic crust during the Ordovician. Consequently, if in the same Variscan internal zone, metabasalts have protolith ages on the one hand at 500 Ma and on the other hand at 470–450 Ma this means that they come from two different oceanic crusts, the youngest being the Rheic Ocean and the oldest the so-called Mirovoi or Ran Ocean, depending on the authors. It is significant to note that similar results were obtained in other segments of the northeastern Variscan belt with similar conclusions (Soejono Žáčková et al., 2010; Collett et al., 2020, 2022).

An important element to emphasize is that even within retrogressed eclogites it is possible to propose quite a well-constrained estimate of the P−T conditions of the metamorphic peak, using mineral inclusions trapped within garnets.

Peak conditions for the MTM eclogites have been estimated at 17.2–18.5 kbar for 640–660 °C. These conditions are consistent with the thermal gradients calculated for the subduction of the mature ocean floor (i.e., “standard subduction zones”, Fig. 15).

By comparison with other available P−T estimates within the MECS microplate (Lardeaux et al., 1994; Cortesogno et al., 2004; Franceschelli et al., 2005; Giacomini et al., 2005; Cruciani et al., 2011, 2015), our results are clearly consistent with most of the P−T values proposed for Sardinian eclogites (see Supplementary material, Table S5).

Altogether these results suggest that the P−T conditions obtained on the severely retrogressed Corsican or on some Sardinian eclogites (SA-4 and Co boxes in Fig. 14) probably reflect postpeak thermal re-equilibration during continental collision. On the other hand, the geochronological data obtained from our samples do not allow us to constrain the age of the eclogite-facies metamorphism in the MTM. It is only possible to state that it predates the amphibolite-facies regional metamorphism for which the older ages range from 360–350 Ma (Oliot et al., 2015). Similarly, there is no robust geochronological constraint for the age of eclogite-facies metamorphism in Corsica or in Sardinia. In Corsica, an age of 361 ± 3 Ma is proposed for HP granulite-facies (Giacomini et al., 2008). Metamorphic evolution under lower pressure conditions is dated between ca. 360 and 330 Ma and described in the Corsican basement (Li et al., 2014; Massonne et al., 2018; Cruciani et al., 2020). In Sardinia, eclogite-facies metamorphism developed prior to the regional-scale Barrovian-type metamorphism that has been dated to ca. 350 Ma (del Moro et al., 1991; Franceschelli et al., 2007; Carosi et al., 2012).

In conclusion, the new data presented in this study show that following an adapted petro-chronological approach it is possible, even in severely retrogressed eclogites, to obtain information on the age of the magmatic protolith and the peak P−T conditions. The conditions of eclogite-facies metamorphism are quantified for the first time in the MTM (17.2–18.5 kbar and 640–660 °C) and are compatible with thermal gradients typical for the subduction of an oceanic crust. The integration of our data on the nature and age of the protoliths in the recently proposed unified full plate paleaomagnetic reconstruction models suggests that the MTM eclogites are derived from the subduction of an oceanic domain older than the Rheic Ocean, located (at ca. 500 Ma) to the north of the Gondwana active margin.

This work was supported by the Research Project no. 310560 of the Czech Geological Survey (DKRVO/ČGS 2018–2022). We are grateful to M. Poujol (Geosciences Rennes, France), N. Novotná and J. Míková (Czech Geological Survey, Prague) for the LA-ICP-MS analyses. We also thank M. Bonnefoy and M. Štrba for mineral separation. We sincerely would like to thank the reviews, the comments and the suggestions of two anonymous reviewers that significantly improved the manuscript. We thank O. Vanderhaeghe for helpful editorial work.