Cambro – Ordovician ferrosilicic magmatism along the northern Gondwana margin: constraints from the Cézarenque – Joyeuse gneiss complex (French Massif Central)

– It is well-acknowledged that the northern margin of the Gondwana supercontinent was affected by a major magmatic event at late Cambrian (Furongian) to early Ordovician (Tremadocian – Floian) times. However, an accurate assessment of its extent, origin, and signi ﬁ cance is partly hampered by the incomplete characterization of the numerous gneiss massifs exposed in the inner part of the Variscan belt, as some of them possibly represent dismembered and deformed Furongian – Lower Ordovician igneous bodies. In this study, we document the case of the “ Cézarenque – Joyeuse ” gneiss complex in the Cévennes parautochthon domain of the French Massif Central. The gneisses form decametre-to kilometre-thick concordant massifs interlayered within a pluri-kilometric sequence of mica-and quartz schists. They encompass two main petrological types: augen gneisses and albite gneisses, both typi ﬁ ed by their blue and engulfed quartz grains with the augen facies differing by the presence of centimetre-sized pseudomorphs after K-feldspar and the local preservation of igneous textures. Whole-rock geochemistry highlights that many gneisses have magmatic ferrosilicic (acidic with anomalously high FeO t and low CaO) compositions while others are akin to greywackes. Collectively, it is inferred that the bulk of the Cézarenque – Joyeuse gneisses represents former rhyodacite lava ﬂ ows or ignimbrites and associated epiclastic tuffs. Volumetrically subordinate, ﬁ ner grained, and strongly silicic leucogneisses are interpreted as microgranite dykes originally intrusive within the volcanic edi ﬁ ces. LA – ICP – MS U – Pb dating of magmatic zircon grains extracted from an augen gneiss and a leucogneiss brackets the crystallization age of the silicic magmas between 486.1 ± 5.5 Ma and 483.0 ± 5.5 Ma which unambiguously ties the Cézarenque – Joyeuse gneisses to the Furongian – Lower Ordovician volcanic belt of SW Europe. Inherited zircon date distributions, Ti-in-zircon and zircon saturation thermometry demonstrate that they formed by melting at 750 – 820 ° C of Ediacaran sediments. Zircon Eu/Eu* and Ce/Ce* systematics indicate that the melts were strongly reduced (fO 2 probably close to the values expected for the iron – wüstite buffer), possibly because they interacted during ascent with Lower Cambrian black shales. This would have enhanced Fe solubility in the melt phase and may explain the peculiar ferrosilicic signature displayed by many Furongian – Lower Ordovician igneous rocks in the northern Gondwana realm. We infer that crustal melting resulted from a combination of mantle-derived magma underplating in an intracontinental rift setting and anomalously elevated radiogenic heat production within the Ediacaran sedimentary sequences.

This study focuses on the so-called Cézarenque-Joyeuse gneiss complex exposed in the Cévennes domain of the southeastern French Massif Central (Fig. 1).Ballèvre et al. (2012) postulated that they could represent felsic igneous rocks of the Furongian-Lower Ordovician belt having experienced greenschist-to lower amphibolite-facies metamorphism during the Variscan orogeny.We explore this possibility by reviewing the field relationships, petrography and geochemistry of these gneisses and providing the first LA-ICP-MS zircon U-Pb and trace element data on such lithologies.
Collectively, our results do tie the Cézarenque gneisses to the Furongian-Lower Ordovician magmatic event, provide novel constraints on the petrogenesis of ferrosilicic magmas and offer the opportunity to address the tectonic evolution of this segment of the northern Gondwana margin.

Tectonic-metamorphic evolution of the Cévennes domain
The Cévennes domain is primarily composed of detrital metasediments including a wide variety of petrographic types (mica and quartz schists, locally feldspar-rich, quartzite, calcsilicate gneisses, paragneisses) commonly gathered under the term "Cévennes schists" (Brouder, 1963;Barbey et al., 2015).It also encompasses several gneiss massifs (Peyrolles, Cézarenque-Joyeuse, Fig. 3), all of which forming decametreto (several) kilometre-thick concordant bodies within schists (Brouder, 1963;Weisbrod and Marignac, 1968;Roger, 1969).At the regional scale, four main tectonic-metamorphic events were described.The main deformation phase D 1 produced a near-horizontal schistosity S 1 with a N-S stretching and mineral lineation, a fabric resulting from a combination of pure and simple shear with top-to-the S kinematics (Mattauer and Etchecopar, 1976;Lacassin and van den Driessche, 1983).Ar-Ar dating of micas and amphibole yielded 340-330 Ma dates for D 1 (Caron, 1994) which took place under metamorphic conditions of c. 500 °C and 5.2 kbar (Arnaud, 1997) and would attest to an early crust thickening and nappe stacking stage (Toteu and Macaudière, 1984;Faure et al., 2009b).In the eastern Cévennes, Faure et al. (2009b) advocated that the Cézarenque-Joyeuse gneisses and overlying micaschists represent an allochthonous complex (klippen) stacked over schists during D 1 , putting forward the existence of a highstrain zone at the base of the main gneiss massif (Mattauer and Etchecopar, 1976).This complex was thus regarded as an equivalent of the Lower Gneiss Unit, and the underlying schists attributed to the Parautochthonous Unit sensu Faure et al. (2009a).Alternative views consider that all metamorphic rocks from the Cévennes (including the Cézarenque-Joyeuse gneisses) belong to the Parautochthonous Unit (Matte, 2007).
stamps the transition to an extensional regime which led to the opening of the Alès (Westphalian D-Autunian) coal basin (Allemand et al., 1997).
3 Lithological components of the eastern Cévennes 3.1 The "Cézarenque-Joyeuse" gneiss complex Based on available descriptions from the literature (Brouder, 1963;Elmi et al., 1974Elmi et al., , 1989;;Magontier, 1988) and our own field observations, three lithological facies were identified: augen gneisses, albite gneisses, and leucogneisses.Their descriptions (presented below) are complemented by a review of their whole-rock major element compositions, based on a compilation of available data (Chenevoy, 1968b;Weisbrod, 1970;Magontier, 1988;Elmi et al., 1989) supplemented by two new analyses of an augen gneiss and a tourmaline-bearing leucogneiss (dated samples, see Sect. 4).Their whole-rock chemical compositions were analysed at the Service d'Analyse des Roches et Minéraux (SARM, CNRS, Nancy) from powdered samples using a Thermo Fischer ICap 6500 ICP-OES for major elements and a Thermo Fisher X7 Q-ICP-MS for Zr.The analytical procedures, reproducibility and limits of detection are detailed in Carignan et al. (2001).The compilation and the newly obtained data are provided in the Supplementary Table 1.
In the literature, the term "leptynite" was traditionally used to describe any quartz-feldspar rich (and with < 5 vol.% mica) granofelsic to gneissic lithology and thus encompasses both felsic meta-igneous rocks and impure metasandstones and arkoses.Hence, samples referred to as "leptynites" were screened to retain only those of igneous compositions by: (i) a careful examination of the field descriptions (notably evidence for original intrusive relationships) and, (ii) applying the procedure of Davoine (1969), i.e., selecting the high SiO 2 analyses that met the criteria: (K 2 O þ Na 2 O) > 7% and CaO < 2%.This way, a total of 71 analyses from the previous studies were retained for the three main gneiss facies.In the following, we assume that the whole-rock compositions were not significantly modified during the Variscan metamorphic evolution, as demonstrated for other gneiss massifs of the eastern Massif Central (Couzinié et al., 2017).Geochemical diagrams were plotted using GCDkit (Janou sek et al., 2006).

The augen gneisses
The augen gneisses are strongly to poorly foliated rocks typified by their mm-to cm-sized bluish quartz porphyroclasts generally making up 5-10% of the rock (Figs.4a-4c).Such grains show undulose extinction and are very commonly "corroded" and engulfed (Fig. 5a).Near rectangular (euhedral) to ovoid (lenticular) 1 to 10 cm-large polycrystalline aggregates (Figs.4a-4c; 5-20% of the rock) dominantly comprise albite plus quartz and were demonstrated to represent pseudomorphs after K-feldspar (Chenevoy, 1968b), in agreement with the local preservation of microcline in their inner part (Fig. 5b).Bluish quartz and feldspar aggregates are embedded in a fine-grained (av.0.5 mm) foliated matrix of quartz, albite, muscovite, and biotite (often chloritized).The occurrence of scarce garnet in the matrix along with relictual oligoclase and microcline was reported by Elmi et al. (1989).From a chemical point of view, the augen gneisses show elevated SiO 2 (av.∼68.5 wt.%) and FeO t (av.∼3.9 wt.%), but distinctively low CaO (always < 1.5 wt.%) contents.In the Al/3-Na vs. Al/3-K diagram of de La Roche (1968), designed to discriminate igneous and sedimentary rocks, their compositions largely overlap with those of silicic Variscan plutonic and volcanic rocks of similar SiO 2 contents (Fig. 6a) demonstrating that the protoliths of the augen gneisses were igneous rocks.They show subalkaline rhyodacite compositions (Fig. 6b) and are highly peraluminous (Fig. 6c) which, together with the high Fe and low Ca contents, indicates that the igneous protoliths of the augen gneisses share a ferrosilicic signature (in the sense of Castro et al., 2009).Considering the alkalinity and Fe/Mg balance, the augen gneisses classify as calc-alkaline to alkali-calcic and magnesian (both sensu Frost et al., 2001, Figs. 6d and 6e) with FeO t /(FeO t þ MgO) clustering around 0.7-0.75.Their K/(K þ Na) ratios are always close to 1 (Fig. 6f).

The albite gneisses
The albite gneisses (Fig. 4d) are composed of a foliated fine-grained (< 0.2 mm) matrix of quartz, albite, biotite (often chloritized) and muscovite, with some albite porphyroblasts locally reaching 1-2 mm (Fig. 5c).Muscovite and albite are overall more abundant than in the augen facies.Fragments of ovoid strongly stretched (dismantled) feldspar aggregates along with mm-sized bluish quartz grains locally occur in the foliated groundmass but remain scarce and of smaller size than in the augen gneisses.From a chemical point of view, the compositions of albite gneisses are more varied than those of their augen counterparts (e.g., SiO 2 : 57-72%).A clear overlap exists as illustrated by most geochemical plots but two thirds of the albite gneisses are typified by higher Al-Fe-Mg (Figs. 6a  and 6c) together with lower SiO 2 and alkali contents (Fig. 6b) with respect to the augen facies.Such samples plot in the fields of greywackes (60%) and shales (40%) in the Al/3-Na vs. Al/3-K diagram of de La Roche (1968) and their compositions resemble those of the Cévennes mica schists, albeit differing by more elevated alkali (Na 2 O þ K 2 O) vs. CaO ratios (Fig. 6).

The "Cévennes schists"
This term encompasses a variety of metamorphic rocks derived from a range of pelite, semi-pelite, greywacke, sandstone, and arkose protoliths.The most representative petrographic type is a mica schist showing a well-developed foliation (Fig. 7a) defined by alternating mm-scale quartz and chlorite-muscovite (± graphite ± biotite ± garnet) layers.The mode of quartz can significantly increase and the rocks grade to quartz schists (Fig. 7b).Albite porphyroblasts (0.5-1 mm) are generally scarce (< 15% of the rock) but locally concentrated along decimetre-thick layers (Fig. 7c).Cordierite is restricted to the northern part of the study area (corresponding to the deepest structural levels) and the local occurrence of staurolite was reported in garnet-biotite schists near Ribes (Fig. 3,  Weisbrod, 1970;Harlaux, 2016;Couzinié, 2017).The contour levels were drawn using the kde2d function of R (Venables and Ripley, 2002).The B-A values for sediment-derived experimental melts at 2-15 kbar and 700-900 °C were taken from the compilation in Couzinié et al. (2017).The black arrow in (c) depicts the increase in Fe solubility observed by Gaillard et al. (2001) in a subaluminous 930 °C granitic melt when fO 2 decreases from NNO þ 1.5 down to NNO-0.7 (∼QFM, NNO sanding for Ni-NiO buffer).Bouilhol et al., 2006).A peculiar dark-coloured quartz schist type, composed of fine-grained (sub-mm) quartz (> 80 vol.%), albite and biotite (Fig. 7d), was identified as 50 to 200 m-thick elongated bodies within the Cézarenque-Joyeuse gneisses.

Field relationships between the gneisses and the "Cévennes schists"
At the map scale, the contact between gneisses and the "Cévennes schists" is systematically parallel to the regional foliation and both lithologies appear interlayered.The gneisses define lenticular 10 to 250 m-thick bodies within the schists to the NE (near St-André and Ribes) and a larger massif to the S, reaching a thickness of ∼2 km (Fig. 3).At the outcrop scale, contrasting relationships were observed.The contact is very often gradational, with albite gneisses progressively losing their feldspar load and grading to mica schists.In contrast, SE of Malons (Fig. 3), gneisses and schists are juxtaposed via a shear zone marked by the concentration of quartz lenses, intense folding, and the occurrence of mylonites.Asymmetric folds, interpreted as drag folds, consistently indicate a top-tothe NW transport, i.e., thrusting of the gneisses over the schists (Fig. 4h), in agreement with the observations of Elmi et al. (1989) and Bouilhol et al. (2006).Finally, the eastern contact of the main gneiss massif is underlined by a distinctive decametric white quartzite (quartz > 95% vol.) layer (referred to as the "Peyremale" quartzite; Figs. 3 and 4g).The contact is sharp, parallel to the bedding in the quartzite (evidenced by grain-size variations) and can be followed along several kilometres.
Within the gneiss bodies, the transition between "augen" and "albite" types is very often gradational and intermediate facies do exist (Elmi et al., 1989).Mapping of the augen gneiss-rich zones indicates that those are broadly concordant with the outline of the massif and the regional foliation.Conversely, the transposed contact between the leucogneisses and the augen gneisses, well-exposed in the Baume Valley (Fig. 3), is sharp and discordant (Fig. 4f), in agreement with the observations of Crevola et al. (1983), pointing to an intrusive relationship between the leucogneiss protolith and the augen gneiss protolith.

Zircon U-Pb dating and trace element compositions of the gneisses
Two representative Cézarenque-Joyeuse gneisses were selected for zircon U-Pb dating to clarify the age and origin of their protoliths.Both samples (augen gneiss 18CEZ01, and a tourmaline-bearing leucogneiss 19CEZ54, Figs.4c and 4e) were collected from the Baume river cross-section (see Fig. 3).

Analytical techniques
Zircon grains were separated using standard techniques (jaw crusher, panning, heavy liquids), cast in epoxy resin and polished down to a near-equatorial grain section.Cathodoluminescence imaging was performed at the CRPG (Nancy, France) using a Jeol SM-6510 SEM equipped with a Gatan CL detector.Zircon U-Pb isotope and trace element analyses were carried out at ETH Zürich, Switzerland, by laser ablationinductively coupled plasma-sector field-mass spectrometry using a RESOlution (ASI, Australia) 193 nm ArF excimer laser system attached to an Element XR (Thermo Scientific, Germany) mass spectrometer.Further details on the analytical procedures and the results of secondary standard and sample measurements are available in the Supplementary text and Supplementary Tables 2-5.

Zircon U-Pb dates
For the augen gneiss sample 18CEZ01, a total of 202 analyses were performed on 148 grains, most of which were euhedral, showed pyramidal tips and ranged in length between 150 and 300 mm with aspect ratios from 1:1.8 to 1:3.7.Cathodoluminescence imaging revealed that half of the grains were composed of a (often CL-bright) core wrapped around by a rim showing concentric oscillatory zoning (Fig. 8).Such rims yielded concordant 206  Seven analyses were strongly discordant and not considered further.
For the tourmaline-bearing leucogneiss sample 19CEZ54, 100 analyses were performed on 68 grains which appeared often broken, slightly corroded and of smaller size (130-170 mm, rarely up to 250 mm) than in sample 18CEZ01.Aspect ratios of euhedral unbroken grains were also more variable and generally comprised between 1:2.2 and 1:2.8 but up to 1:5 in some cases.Fifty-two analyses were performed on grains devoid of core-rim relationships but showing prominent concentric oscillatory zoning (striped zoning in the most elongated grains).The concordant analyses clustered between 468 ± 6 (#68) and 489 ± 6 (#90) Ma (n = 24) with 2 grains showing younger dates of 428 ± 12 (#75) and 433 ± 9 (#94) Ma and 12 grains older dates from 545 ± 6 (#44) to 611 ± 9 (#86), up to 1212 ± 6 (#76) Ma.Out of the 16 discordant 206 Pb/ 238 U dates obtained, 8 overlapped with the concordant data (490-480 and at c. 540 Ma) while 8 defined a continuous trend from 455 ± 7 (#87) to 414 ± 6 (#36) Ma.The remaining analyses were performed on grains comprising a relict core surrounded by oscillatory-zoned overgrowths.Eleven rims yielded concordant overlapping 206 Pb/ 238 U dates between   To sum up, both gneiss samples were typified by the occurrence of oscillatory-zoned grains and rims with 206 Pb/ 238 U dates in the range 475-500 Ma, i.e., Furongian-Tremadocian (Fig. 9).The cores and the remaining oscillatoryzoned grains generally displayed Lower Cambrian-Neoproterozoic (525-999 Ma, peak at 600 Ma, Fig. 10b) and Paleoproterozoic-Neoarchean dates (1.8-2.7 Ga).For sample 18CEZ01, the density distribution of the Furongian-Tremadocian dates (n = 72) was clearly symmetrical, centred on 487.4 Ma but importantly, the results of a x 2 red test for homogeneity indicated that the data were overdispersed given the estimated analytical uncertainties (MSWD = 2.3, p-value of 10 À9 ) (Fig. 9).This overdispersion is due to the 5 youngest and 5 oldest dates of the population.Alternatively, the statistical procedure of Montel et al. (1996) shows that the data can be modelled as a mixture of two populations, one centred at 492.2 ± 1.3 Ma (n = 46) and the other at 484.2 ± 1.0 Ma (n = 27, with global p-value of 0.94).The grains belonging to each population are undistinguishable based on texture or trace element compositions.For sample 19CEZ54, 34 analyses clustered in the range 490-468 Ma.Their distribution was asymmetrical and negatively skewed (towards younger ages).Excluding the three youngest analyses allowed to calculate a weighted average date of 483.0 ± 1.2 Ma (± 5.5 after propagation of systematic uncertainties) with an MSWD of 1.46 (p-value = 0.05) indicating that the considered grains may sample a single population.

Zircon trace element compositions
Only compositions obtained from concordant U-Pb spots will be presented and further discussed, summing up to 159 spots for the augen gneiss 18CEZ01 and 64 for the leucogneiss 19CEZ54.Eu/Eu* were calculated as Eu N / (Sm N Â Gd N ) 0.5 but Ce/Ce* as (Nd N ) 2 /Sm N following the approach of Loader et al. (2017).Since this method does not rely on the La and Pr contents, it is less sensitive to presence of minute inclusions of apatite, monazite or xenotime which would flaw the LREE budget of the analysed zircon grain and lead to inaccurate estimates of Ce/Ce* (e.g., Ni et al., 2020).In the formulas, the N stands for "normalized to the chondritic values", here of Boynton (1984).Ranges given in the description below correspond to the 1st and 3rd quartiles of the distribution unless stated otherwise.
A similar situation was observed in leucogneiss sample 19CEZ54 (Fig. 11) with all grains exhibiting steep HREE patterns (with (Dy/Yb) N centred around 0.16-0.25)and variable Ti contents (5.4-10.3 ppm)  The background yellow shading mimics the contours of the distribution of 209 zircon analyses from S-type granites which whole-rock compositions are not ferrosilicic (data from Wang et al., 2012: Gao et al., 2016: Burnham and Berry, 2017), drawn using the kde2d function of R (Venables and Ripley, 2002).For sake of consistency, Ce/Ce* of the literature zircon were recalculated using the same methodology as for the Cézarenque-Joyeuse gneisses, i.e., following the approach of Loader et al. (2017).Were also plotted the relationships between zircon Eu and Ce anomalies and melt temperature-oxygen fugacity, as estimated by Trail et al. (2012) for peraluminous melt compositions.The oxygen fugacity (fO 2 ) in this plot is expressed relative to the NNO (nickel-nickel oxide) buffer.At 800 °C, the fO 2 of the quartz-fayalite-magnetite (QFM) buffer corresponds to NNO-0.8 and that of the iron-wüstite buffer is NNO-5.

Ti-in-zircon and zircon saturation thermometry
The crystallization temperatures of Furongian-Tremadocian zircon grains and rims were calculated from their Ti contents using the equation of Ferry and Watson (2007) with a SiO2 of 1 and a TiO2 of 0.5 as suggested by Schiller and Finger (2019) for S-type felsic magmas.In both samples, estimated values spread over a large range of temperatures (between 670 and 900 °C) but with a main cluster at 730-770 °C (Fig. 12).
Zircon saturation temperatures were obtained based on the major element and Zr contents of the dated samples using the equation of Watson and Harrison (1983).A melt having a composition akin to augen gneiss sample 18CEZ01 would be saturated in zircon below 844 ± 42 °C (considering a 5% absolute uncertainty).A lower temperature of 721 ± 36 °C was retrieved for the leucogneiss sample.Given the common preservation of zircon cores in both samples, these values should be regarded as maximum magma temperatures.

Discussion
5.1 Nature and age of the protoliths of the Cézarenque-Joyeuse gneisses 5.1.1Interpretation of the field relationships, petrography, and whole-rock compositions Three models exist regarding the origin of the Cézarenque-Joyeuse gneisses: (a) they represent metasediments (conglomerates and arkoses) reworking materials of igneous origin (granite pebbles, microcline clasts, Weisbrod and Marignac, 1968); (b) they are metagranites (Crevola et al., 1983); (c) they correspond to volcanic and volcano-sedimentary rocks (Chenevoy, 1968a(Chenevoy, , 1968b;;Faure et al., 2001).Model (a) should be discounted as a significant proportion of the gneisses (the augen facies and at least a third of the albite gneisses) show acidic igneous compositions (see Sect. 3.1).Model (c) is preferred against model (b) as the latter cannot explain the gradual transition at the outcrop scale from rocks of igneous vs. sedimentary (two thirds of the albite gneisses) origin.The volcanic model is also supported by the elevated normative Qz/ (Ab þ Or) ratios (average of 0.66) of the igneous samples consistent with a low pressure of crystallization of their parental magmas (Wilke et al., 2017) and the sharp, stratigraphic, contact between the gneisses and the overlying Peyremale quartzite (see Sect. 3.3), pointing to a near-surface emplacement.
In this frame, based on the SiO 2 -alkalis systematics (Fig. 5b), we infer that the augen and albite gneisses with igneous compositions represent now-metamorphosed rhyodacites.Their blue quartz grains and pseudomorphs after K-feldspar (for the augen facies) embedded in a finer grained matrix are regarded as former igneous phenocrysts, a view consistent with the fact that the blue colour of quartz is commonly interpreted as resulting from rutile exsolution from grains crystallized at high temperature (T > 700 °C, Seifert et al., 2011).The persistence of blue quartz grains in the albite gneisses of sedimentary origin and their composition akin to greywackes indicate that they represent the proximal remobilization products of the adjacent rhyodacites, i.e., epiclastic tuffs.Finally, the leucogneisses are strongly silicic and show a large range of Fe/Mg ratios (Fig. 6) which are typical of fractionated granitic melts.Given their cross-cutting field relationships and their elevated normative Qz/(Ab þ Or) ratios (again pointing to crystallization at low pressure), they likely represent former microgranite dykes intrusive within the rhyodacites and associated epiclastic tuffs.

Interpretation of U-Pb results
In line with the abovementioned arguments, the augen gneiss sample 18CEZ01 should be regarded as a metamorphosed porphyritic (K-feldspar-bearing) rhyodacite and the leucogneiss sample 19CEZ54 as a metamicrogranite.In the following, only concordant zircon analyses will be discussed, discordant results being ascribed to a combination of common Pb incorporation and Pb loss.
The augen gneiss (metarhyodacite) 18CEZ01 contains many magmatic grains and rims yielding Furongian-Tremadocian 206 Pb/ 238 U dates, the distribution of which is wellcentred around 487 Ma and would correspond to the crystallization age of the magma.Importantly, considering the estimated analytical uncertainties, the dates are statistically overdispersed.The lack of correlation between U contents and 206 Pb/ 238 U dates and the symmetrical shape of the date distribution collectively suggest that this overdispersion is of geologic significance (Spencer et al., 2016) and should not be ascribed to limited Pb loss.In line with the volcanic origin of 18CEZ01, it is likely that the analysed dataset comprises syneruptive zircon grains and rims plus antecrysts which formed in deep-seated magma chambers (e.g., Matzel et al., 2006) and were scavenged during the eruption.For such cases (mixture of pre-eruptive and syn-eruptive grains, undistinguishable on textural and chemical grounds), Vermeesch (2021) argued that the "Maximum Likelihood Age" (MLA) model designed by Galbraith (2005) has a high potential to unravel the actual crystallization age of the igneous rock.Indeed, the MLA model presumes that the antecrysts define a continuous range of preeruptive dates, which is arguably more realistic than the "two populations" (i.e., two age clusters) subdivision (see Sect. 4.2) inferred based on the statistical procedure of Montel et al. (1996).Running the MLA algorithm yields 486.1 ± 0.9 Ma (± 5.5 Ma when systematic uncertainties are considered), which is the best estimate of the eruption age of the augen gneiss parental magma.We interpret the grains and cores showing older dates (from the Lower Cambrian to Paleoarchean) as inherited from the magma source or as xenocrysts incorporated from the country-rocks during magma ascent and emplacement (Fig. 9).
For the metamicrogranite 19CEZ54, most oscillatoryzoned grains and rims also yielded 206 Pb/ 238 U dates in the range 475-500 Ma with two concordant analyses showing overlapping younger dates of 428 ± 12 Ma (#75) and 433 ± 9 Ma (#94).We do not ascribe any significance to these dates because in both cases the 206 Pb/ 238 U date / 207 Pb/ 206 Pb date ratios are < 90%, hence suggesting postmagmatic disturbance of the isotopic system (Spencer et al., 2016).Besides, the grain domain corresponding to analysis #94 showed elevated Ti (∼160 ppm), Nb-Ta (9 and 10 ppm, respectively) and Pr contents (0.8 ppm), which are an order of magnitude higher than in other zircon grains and likely attests to the presence of zirconolite, a common product of zircon alteration (Gieré, 1996).Excluding these two dates, we infer that the cluster at c. 483 Ma represents the crystallization age of the parental magma, in agreement with textural evidence.Importantly, the negatively skewed distribution argues for the occurrence of Pb loss from the 475-500 Ma magmatic population (Spencer et al., 2016) and, excluding the 3 youngest dates, we retain the weighted average date of 483.0 ± 1.2 Ma (± 5.5 Ma considering systematic uncertainties) as the best estimate of the crystallization age of the microgranite protolith.
The difference of 3 Ma between zircon crystallization ages in the two felsic magmas exceeds the internal uncertainties and, since both samples were dated during the same analytical session, such time interval may be of geological significance (Horstwood et al., 2016).The slightly younger crystallization age of the microgranite 19CEZ54 is consistent with field observations supporting an intrusive relationship with the porphyritic metarhyodacite 18CEZ01.

Typology and petrogenesis of the igneous association
Nearly overlapping crystallization ages and key geochemical markers (including almost identical whole-rock Na/K ratios, Fig. 6d, and similar magmatic zircon trace element systematics, Fig. 11) collectively indicate that the metarhyodacites and the metamicrogranites were genetically related and constitute a single magmatic system.Considered together, they define a peraluminous, calc-alkalic to alkali-calcic, magnesian and sodi-potassic association (Fig. 6), which is the hallmark of crust-derived magmatic suites (Bonin et al., 2020).Further evidence for a crustal origin is provided by the marked zircon inheritance (Laurent et al., 2017) and zircon trace element systematics.Indeed, the low Th/U ratios displayed by the magmatic zircon grains, positively correlated with LREE contents (Figs.11e and 11f), can be regarded as a consequence of coeval monazite precipitation from the melt upon cooling which is a typical feature of highly aluminous and P-rich "S-type" granitic melts formed by melting of sedimentary rocks (Cuney and Friedrich, 1987).Besides, as Ce 4þ and Eu 3þ are less incompatible with respect to zircon than Ce 3þ and Eu 2þ (e.g., Trail et al., 2012, and references therein), the low Ce/Ce* and Eu/Eu* ratios (weak Ce positive anomalies and marked Eu negative anomalies, Figs.11g and 11h) displayed by magmatic zircon grains demonstrate that the parental magmas were notably reduced, as commonly observed in S-type granitic melts (Whalen and Chappell, 1988).
Examination of the inherited zircon date distributions and trace element systematics sheds light on the nature and origin of the crustal source.The main Neoproterozoic

Leucogneiss 19CEZ54
Furongian-Tremadocian grains and rims; n=34 Interestingly, the inherited Precambrian grains have a wider range of trace element compositions compared to magmatic grains and include a significant proportion of zircon with higher Th/U and LREE contents together with stronger positive Ce anomalies and weaker Eu negative anomalies.These latter signatures suggest that the Precambrian igneous rocks which erosion products fed the Ediacaran sedimentary basins largely originated from more oxidised (Figs.11g and  11h), probably arc-derived magmas (Ishihara, 2004).For the Neoproterozoic grains, the long-lived accretionary Cadomian orogeny, developed along the northern Gondwana margin (Garfunkel, 2015, and references therein), constitutes the most likely source.The Paleoproterozoic to Archean zircon grains would stem from the recycling of old Gondwanan crustal materials, possibly originating from the Saharan Metacraton (Couzinié et al., 2019).
The Cézarenque-Joyeuse felsic igneous rocks have wholerock compositions matching those of the so-called ferrosilicic suites (Castro et al., 2009), i.e., they are anomalously rich in Fe and Mg and low in Ca.Such signatures have been thought to result from very high melting temperatures (> 1000 °C) of sedimentary rocks (mostly greywackes, Castro et al., 2009).However, several observations argue against this model for the Cézarenque-Joyeuse gneisses.First, a large pool of inherited zircon grains and cores were preserved throughout the magmatic evolution implying either moderate melting temperatures (< 850 °C, below the zircon saturation temperature of the augen gneiss) or very fast melt production and transfer to the upper crust (Watson, 1996;Bea et al., 2007).In the latter case, the magmas are expected to have followed a nearly adiabatic path from source to surface.This way, crystallization temperatures deduced from Ti-in-zircon thermometry (clustering at 730-770 °C, see Sect.4.4) should lie within 50 °C of the actual melting temperatures (Holtz and Johannes, 1994), which, therefore, could not have exceeded 820 °C.As an alternative to the very high melting temperature model, Fiannacca et al. (2019) suggested that the ferrosilicic signature should be explained by the selective incorporation in the melt phase of mafic materials (mostly peritectic garnet, Stevens et al., 2007) present in the sedimentary source.Yet appealing, this model cannot be directly tested in that any petrographic evidence for such entrainment was irremediably erased during the dissolution of the added crystal load.
Novel insights on the origin of the ferrosilicic signature may be gained from the examination of the Eu/Eu* and Ce/Ce* systematics of the Cézarenque-Joyeuse magmatic zircon grains and rims.Measured values (Figs.10g and 10h) are significantly lower compared to zircon encountered in "typical" (not ferrosilicic) sediment-derived granites from the Lachlan Fold Belt (Burnham and Berry, 2017) and southern Tibet (Wang et al., 2012;Gao et al., 2016), calling for anomalously reducing conditions.The fO 2 prevailing during zircon crystallization was most likely close to the value expected for the iron-wüstite buffer considering the calibration of Trail et al. (2012).Little is known on the influence of fO 2 on anatectic melt compositions, but Gaillard et al. (2001) showed that the Fe solubility of a subaluminous (A/CNK between 1 and 1.1) melt at 930 °C is negatively correlated to fO 2 : at NNO þ 1.5, maximal FeO t contents reach ∼1.8 wt.% but are higher than 3 wt.% at fO 2 below FMQ.Since the alumina content of a granite melt at a given fO 2 has no influence on the Fe solubility (Holtz et al., 1992), this result should also be valid for peraluminous melt compositions.Hence, we posit that the high Fe contents displayed by the ferrosilicic rocks may result from an enhanced Fe solubility in the melt phase itself caused by strongly reducing conditions.Those may be inherited from the source level (melting in a graphite-buffered environment) or acquired during ascent via interaction with organic matter (possibly oil-bearing) sedimentary rocks, a phenomenon known to deeply affect the fO 2 of magmas (e.g., Iacono-Marziano et al., 2012).As a matter of fact, the Lower Cambrian sequences of the southern Massif Central autochthon host a > 200 m-thick black shale formation (Álvaro et al., 2014), to which graphite-bearing mica schists of the Cévennes parautochthon may be correlated.Hence, the low fO 2 and peculiar high Fe contents of the Cézarenque-Joyeuse parental magmas may result from their interaction with the local Lower Cambrian sediments prior to the eruption (Fig. 13).

5.3
The Cézarenque-Joyeuse magmatism in the frame of the northern Gondwana evolution Felsic (sub)volcanic and volcanosedimentary associations coeval to the Cézarenque-Joyeuse gneisses are widespread over the northern Gondwana terrains of SW Europe.In this section, we first provide a short review of correlative formations and then address the geodynamic setting and the trigger(s) of this magmatic flare-up.
Altogether, petrological and geochronological evidence substantiate that a pervasive crustal melting event took place along a > 2000 km segment of the northern Gondwana margin (Álvaro et al., 2020a, b).From a geodynamic perspective, an intracontinental rift setting (Fig. 13) should be retained based on: (i) the linear shape of the magmatic belt and the synchronous deposition of thick sedimentary sequences (Pouclet et al., 2017); (ii) the occurrence of "anorogenic" A-type magmatic rocks in NW Iberia (Díez Fernández et al., 2012, and references therein), in the Maures massif (Seyler, 1986;Briand et al., 2002), in the Armorican Massif (Ballèvre et al., 2012) and in the southern Massif Central (Albigeois area, Pin and Marini, 1993); (iii) the lack of regional metamorphism related to crust thickening (e.g., Montero et al., 2007); and (iv) the coeval opening of the Rheic ocean (Díez Montes et al., 2010;Nance et al., 2010).The ultimate origin of this rifting event remains debated.Many authors put forward the role of the southwards (in Ordovician coordinates) subducting Iapetus slab in controlling crustal extension (Fernández et al., 2008;Díez Montes et al., 2010;Díaz-Alvarado et al., 2016;Oriolo et al., 2021) while others retained a plume-induced origin (Briand et al., 1992(Briand et al., , 2002)).The geological record of the Cévennes parautochthon does not provide any evidence that would help clarify this point.
An atypical feature of this rifting event is the voluminous generation of peraluminous crust-derived magmas, which are more commonly encountered in syn-to post-collisional orogenic settings (Barbarin, 1999).Two factors seem to have played a key role in enabling the crust to produce granitic melts under such conditions.First, evidence for mantle-derived magma underplating and associated advective/conductive heat transfer are provided by the occurrence of coeval mafic magmatic rocks in the vicinity of the felsic volcanic centres (Bea et al., 2007;Montero et al., 2009;Díez Montes et al., 2010;Díaz-Alvarado et al., 2016;Pouclet et al., 2017).However, such rocks represent small volumes compared to the felsic suites and are not systematically exposed (e.g., in the Cézarenque-Joyeuse area) thus calling for an additional heat supply.As a matter of fact, the thick Ediacaran sedimentary sequences which constitute the source rocks of the Furongian-Lower Ordovician magmas were characterized by anomalously high radiogenic heat productions: > 2.7 mW.m À3 at 550 Ma in the Iberian "Schist and Greywacke Complex" (Bea et al., 2003) and average of 3.0 mW.m À3 in the Massif Central (Couzinié, 2017), i.e., 50% higher than the mean upper crustal composition of Rudnick and Gao (2003).This feature was explained by the recycling of Cadomian felsic igneous rocks which selectively enriched the Ediacaran detritus in K, Th and U (Bea et al., 2003).Numerical modelling indicates that the presence of layers with elevated radiogenic heat productions within the crust can significantly affect its thermal structure and provoke heating during subsidence (Sandiford et al., 1998).Therefore, we infer that the pre-rifting crust structure and composition (inherited from its late Neoproterozoic evolution) played an important role in enhancing melt production at Furongian-Lower Ordovician times.

Conclusion
Field relationships, petrography and whole-rock geochemistry collectively indicate that the Cézarenque-Joyeuse gneisses represent former volcanic rocks of rhyodacite composition and their erosion products, originally interlayered within detrital sedimentary sequences.Zircon U-Pb dating of the gneisses demonstrate that the felsic magmas erupted or were emplaced at very shallow crustal levels between 486.1 ± 5.5 Ma and 483.0 ± 5.5 Ma.In that regard, the Cézarenque-Joyeuse gneisses do represent a newly identified fragment of the Furongian-Lower Ordovician volcanic belt of SW Europe.Inherited zircon date distribution and magmatic zircon trace element systematics further substantiate that the source of the Furongian-Lower Ordovician magmas corresponded to Ediacaran sediments and suggest that the ferrosilicic signature they commonly exhibit may be explained by the strongly reduced character of the silicic melts (fO 2 close to the iron-wüstite buffer) acquired through interaction with Lower Cambrian organic matter-bearing sediments (Fig. 13).Such a model would be valid for other Furongian-Lower Ordovician ferrosilicic rocks of the northern Gondwana realm as Lower to Middle Cambrian black shales have been described in the Central Iberian Zone (Álvaro et al., 2020a, b) and in the southern Armorican massif (Pouclet et al., 2017).In light with available paleogeographic constraints, it is inferred that crustal melting took place in an intracontinental rift setting and was enhanced by mantle-derived magma underplating and the anomalously high radiogenic heat production of the Ediacaran sedimentary sequences.

Fig. 3 .
Fig.3.Geological map and associated interpretative cross-sections of the eastern Cévennes highlighting the intricate association between the different gneiss facies and the "Cévennes schists".The location of the dated samples is indicated in the inset and the cross-section centred on the Baume valley.Map redrawn and adapted from the works ofElmi et al. (1974Elmi et al. ( , 1989) ) andBouilhol et al. (2006).

Fig. 8 .
Fig. 8. Representative cathodoluminescence images of zircon grains from the Cézarenque gneisses.The locations of laser spots (white circles) are indicated along with the spot name (#XX).The corresponding 206 Pb/ 238 U dates (if < 1.2 Ga) or 207 Pb/ 206 Pb dates (if > 1.2 Ga) are quoted with 2s uncertainty, in Ma.All displayed analyses are concordant.

Fig. 11 .
Fig. 11.Zircon trace element data for the Cézarenque-Joyeuse gneisses.(a,b,c,d) Rare Earth Elements patterns normalized to the chondrite values of Boynton (1984); (e,f) LREE contents (Ce þ Nd) vs. Th/U diagram for the different date populations; (g,h) Eu/Eu* vs. Ce/Ce* diagram.The background yellow shading mimics the contours of the distribution of 209 zircon analyses from S-type granites which whole-rock compositions are not ferrosilicic (data fromWang et al., 2012: Gao et al., 2016: Burnham and Berry, 2017), drawn using the kde2d function of R(Venables and Ripley, 2002).For sake of consistency, Ce/Ce* of the literature zircon were recalculated using the same methodology as for the Cézarenque-Joyeuse gneisses, i.e., following the approach ofLoader et al. (2017).Were also plotted the relationships between zircon Eu and Ce anomalies and melt temperature-oxygen fugacity, as estimated byTrail et al. (2012) for peraluminous melt compositions.The oxygen fugacity (fO 2 ) in this plot is expressed relative to the NNO (nickel-nickel oxide) buffer.At 800 °C, the fO 2 of the quartz-fayalite-magnetite (QFM) buffer corresponds to NNO-0.8 and that of the iron-wüstite buffer is NNO-5.

Fig. 12 .
Fig.12.Histograms and density distribution of temperatures for magmatic Furongian-Lower Ordovician grains and rims of the Cézarenque-Joyeuse gneisses, obtained via the Ti-in zircon thermometer using the equation of Ferry andWatson (2007) with a SiO2 of 1 and a TiO2 of 0.5 as suggested bySchiller and Finger (2019) for peraluminous felsic magmas.

Fig. 13 .
Fig. 13.Interpretative geodynamic sketch illustrating the Furongian-Lower Ordovician evolution of the northern Gondwana margin.The crustal column on the right panel summarises the petrogenetic model proposed for the protoliths of the Cézarenque-Joyeuse gneiss complex.