3.1. Sampling

Eight samples have been collected in the entire basin (Figures 1A and 2 and Table 1). Three of them (Per21-3, Per21-4 and Can21-5) have been sampled in the so-called Cinéritique Fm belonging to the Stephanian subdivision, but due to lack of continuous outcrops, their precise stratigraphic location in this Fm is not known (Figures 1A and 2). One sample (Dev21-1) was collected in the Saint-Rome-de-Tarn Fm (F2), in the Grey Autunian subdivision (Figures 1A, B and 2). Another one (Lat20-1) is located within the Dourdou Fm (F3; Figures 1A, B, 2 and 3A) and corresponds to Rolando’s cinerite III (1988). The remaining three samples (Gal21-6, Gal21-7 and Cam20-3) were collected within the Saint-Pierre transitional facies Fm (F4/F5), close to the top of the red Autunian subdivision (Figures 1A, B, 2 and 3B for sample Cam20-3).

3.2. Whole-rock geochemical analyses

Samples were crushed at Thin Section Lab (TSL, Toul, France) following a standard protocol to obtain adequate powder fractions using agate mortars. Chemical analyses (Supplementary Table 1) were performed by the Service d’Analyse des Roches et des Minéraux (SARM; CRPG-CNRS, Nancy, France) using an ICP–AES (inductively coupled plasma–atomic emission spectroscopy) for major-elements and an ICP–MS (inductively coupled plasma–mass spectrometry) for trace-elements, following the techniques described in Carignan et al. [2001]. All geochemical classification and tectonic discrimination diagrams used in this study were drawn using the GCDkit software [Janousek et al. 2006].

3.3. Zircon U–Pb dating

A classic mineral separation procedure has been applied to concentrate zircon grains suitable for U–Pb dating using the facilities available at TSL. Zircon grains were imaged by cathode luminescence (CL) using a Reliotron CL system equipped with a digital colour camera available at the GeOHeLiS platform (University of Rennes 1, France).

U–Pb geochronology of zircon was conducted by in-situ LA–ICP–MS (laser ablation–inductively coupled plasma–mass spectrometry) at the GeOHeLiS analytical platform using an ESI NWR193UC Excimer laser (193 nm wavelength), coupled to a quadripole Agilent 7700x ICP–MS equipped with a dual pumping system to enhance sensitivity [Paquette et al. 2014]. The methodology used to perform the analyses can be found in Nosenzo et al. [2022] as well as in Supplementary Table 2. Concordia diagrams have been generated using IsoplotR [Vermeesch 2018].

3.4. Zircon Hf analyses

Lu–Hf isotopes were measured at the Géosciences Montpellier Laboratory, University of Montpellier (AETE–ISO regional facility of the OSU OREME) using a ThermoFinnigan Neptune+ MC–ICP–MS (multicollector–inductively coupled plasma–mass spectrometer) coupled with a Photon-Machine Analyte G2 Excimer laser (193 nm wavelength). Ablation was performed using a 50 μm spot size. The laser frequency was 5 Hz and the energy density of the laser beam was c. 6 J∕cm². A typical analysis was 80 s, including a 40 s background measurement and a 40 s ablation period. The correction for the interferences and mass bias followed the procedure outlined in previous reports [e.g. Bruguier et al. 2020]. The correction for the isobaric interference of Yb and Lu on ¹⁷⁶Hf was made following a method detailed in Fisher et al. [2011]. For Yb, the interference-free ¹⁷¹Yb was corrected for mass bias effects using an exponential law and ¹⁷³Yb∕¹⁷¹Yb = 1.130172 [Segal et al. 2003]. The mass bias-corrected ¹⁷¹Yb was monitored during the run and the magnitude of the ¹⁷⁶Yb interference on ¹⁷⁶Hf was calculated using ¹⁷⁶Yb∕¹⁷¹Yb = 0.897145 [Segal et al. 2003]. For Lu, the interference-free ¹⁷⁵Lu was corrected for mass bias effects assuming βLu = βYb and using an exponential law. The mass bias-corrected ¹⁷⁵Lu was monitored during the run and the magnitude of the ¹⁷⁶Lu interference on ¹⁷⁶Hf was calculated using ¹⁷⁶Lu∕¹⁷⁵Lu = 0.02655 [Vervoort et al. 2004]. Interference-corrected ¹⁷⁶Hf∕¹⁷⁷Hf were corrected for mass bias using an exponential law and ¹⁷⁹Hf∕¹⁷⁷Hf = 0.7325 [Patchett et al. 1981]. Initial Hf isotope ratios and εHf values were calculated using the decay constant for ¹⁷⁶Lu of 1.867 × 10⁻¹¹ yr⁻¹ [Söderlund et al. 2004] and the CHUR values of 0.282785 and 0.0336 for ¹⁷⁶Hf∕¹⁷⁷Hf and ¹⁷⁶Lu∕¹⁷⁷Hf [Bouvier et al. 2008]. The accuracy and long-term reproducibility of the measurements were gauged by analysing three zircon reference standards [91,500, GJ1 and Plešovice with reference values taken from Blichert-Toft 2008; Morel et al. 2008; Sláma et al. 2008, respectively] and all values were found to be in agreement with the reference values: 91,500 (¹⁷⁶Hf∕¹⁷⁷Hf = 0.282298 ± 24, n = 30), Plešovice (¹⁷⁶Hf∕¹⁷⁷Hf = 0.282473 ± 19, n = 13) and GJ1 (¹⁷⁶Hf∕¹⁷⁷Hf = 0.282007 ± 32, n = 15) (all errors at 2 s.d. level).

4.1. Petrology

Sample Per21-4, from the Cinéritique Fm (Figures 1A and 2), is a tuffite characterized by a centimetric-scale layering related to grain size alternation. In thin section, it is dominated by a very fine-grained matrix including floating flat parallel unsorted silt to rare coarse sand-size clasts responsible for the laminated fabric. The clast fraction is mainly composed of large elongated quartz grains including micro bubble-like cavities compatible with ancient glass shards, K-feldspars and abundant muscovite flakes (Figure 4A). Secondary carbonates can also be observed.

Also from the Cinéritique Fm (Figure 2), the matrix of sample Can21-5 was collected in a 20 cm thick light-beige layer containing numerous plant remnants. It is extremely fine-grained and contains a dispersed very-fine silt-size fraction, including resorbed quartz crystal clasts (Figure 4B) and K-feldspar clasts, consistent with a pyroclastic origin, as well as terrigenous muscovite flakes.

Sample Dev21-1 corresponds to a 5 cm thick very fine-grained light-grey layer belonging to the Saint-Rome-de-Tarn Fm (Figures 1A and 2). In thin section (Figure 4C), it is composed of an extremely fine-grained matrix-supported silt- to sand-size unsorted clastic fraction, including larger angular to subangular quartz fragments of possible altered glass shards, K-feldspar clasts, as well as rare tiny terrigenous muscovite flakes, indicating the lack of hydrodynamical sorting (Figure 4C).

Sample Lat20-1 belongs to the Dourdou Fm (F3 Unit, Figures 1A, 2 and 3A). It is a pinkish to light-beige fine-grained tuff with an eutaxitic-like fabric marked by black flame-like structures, suggesting a pyroclastic flow deposit, rather than a fall deposit. In thin section (Figure 4D), the sample is highly recrystallized and displays an anhedral quartz-rich mosaic invading the residual cryptocrystalline matrix with large dispersed carbonate patches and remaining K-feldspar and plagioclase clasts. A striking feature is the strong late carbonatation and recrystallization that this tuff suffered, possibly during the diagenetic processes.

Sample Gal21-7 belongs to the Saint-Pierre Fm (F5 Fm, Figure 2). In thin section, it is characterized by a dominant cryptocrystalline matrix, probably at least partly vitreous, containing millimetre thick bands enriched with fine sand-size to silt-size quartz and K-feldspar debris, together with smaller biotite and detrital muscovite grains. As shown on Figure 4E, some of the quartz clasts show resorption gulfs, which is consistent with their interpretation as rhyolitic quartz that underwent fast decompression in volcanic conduits. Lithic fragments, among which angular polycrystalline quartz clasts, can also be found.

Samples Cam20-3 and Gal21-6 were sampled in the same 20 cm thick tuffitic layer of the Saint-Pierre Fm (F5, Figures 1A, 2 and 3B). Sample Cam20-3 is a light-grey tuffite with a centimetric scale layering. In thin section, it contains a fair number of lithic fragments and angular quartz clasts (Figure 4F) compatible with a pyroclastic origin, as well as K-feldspar and plagioclase clasts together with chlorite after biotite, muscovite, accessory tourmaline and opaque minerals into a cryptocrystalline matrix. Muscovite and tourmaline indicate a mixing of the pyroclastic content with a terrigenous input. As shown on Figure 4F, this sample also reveals the presence of secondary carbonate minerals. Sample Gal21-6 is a very fine-grained tuffite similar to sample Gal21-7, but with a much smaller grain size with a clear planar microfabric marked by the alignment of the quartz clasts and muscovite microdebris (Figure 4G). Some late goethite veins can be found in this sample (Figure 4G).

4.2. Whole-Rock geochemistry

Eight samples have been selected for major and trace element analyses (Supplementary Table 1). As illustrated in the previous section, all samples consist of ash tuffs and tuffites in which the terrigenous input is variable but obviously difficult to accurately quantify for each of the samples, a feature that may disturb their volcanic geochemical signatures. In addition, some samples underwent late (i.e. post-deposition) diagenetic modifications, such as carbonatation, sometimes fairly important, as for sample Lat20-1 that contains more than 10 wt% CaO. This increase in CaO (and Sr) content consequently dilutes the other elements. These characteristics can limit the usefulness of geochemical diagrams, which therefore need to be used with caution, especially those based on major element concentrations. In the conventional total alkali-silica (TAS) diagram (Figure 5A), sample Lat20-1 plots in the field of basaltic trachy-andesite, while all the others plot in the rhyolite field.

Whole-rock immobile trace and REE elements are often used in the case of volcanic ashes studies as they are usually less sensitive to post deposition processes (such as diagenesis). In the Zr/TiO 2 versus Nb/Y diagram of Winchester and Floyd [1977], two distinct groups appear (Figure 5B). A first group, comprising Per21-4, Per21-3, Can21-5 and Dev21-1 plots in the field of trachyte/trachy-andesite whereas a second group (Gal21-6, Gal21-7, Cam20-3 and Lat20-1) plots in the rhyodacite/dacite field. It should be noticed that yttrium can be lost during the alteration of volcanic rocks [Hill et al. 2000] and therefore the alkalinity attributed to the samples should be considered as a maximum estimate.

In a Primitive Mantle normalized multi-elements diagram [McDonough and Sun 1995, Figure 6a], all the studied samples yield trends that are comparable with the Upper Continental Crust [UCC, Taylor and McLennan 1985]. Overall, the samples show a strong enrichment in LREE (Figure 5B), and, with the exception of sample Dev21-1, a negative Eu/Eu* anomaly, often found in felsic volcanic rocks [Wray 1999]. In further detail, it is noteworthy that samples belonging to the two groups defined above present two distinct trends overall (Supplementary Table 1 and Figure 6B). Samples from the first group (Per21-4, Per21-3, Can21-5 and Dev21-1) display a low REE content (22 to 98 ppm), a less pronounced negative Eu/Eu* anomaly (0.65 to 0.98) and strong HREE depletions. In contrast, samples from the second group (Gal21-6, Gal21-7, Cam20-3 and Lat20-1) present a higher REE content (143 to 158 ppm), a more pronounced Eu/Eu* anomaly (between 0.17 to 0.69) and flat HREE patterns. If the REE contents are normalized to that of the Cody shale [Figure 6C; Jarvis and Jarvis 1985], most samples show a negative Eu anomaly, are depleted in LREE and are either flat (second group) or depleted (first group) in HREE, which are all typical features for volcanic ashes [Wray and Wood 1998; Ducassou et al. 2019]. The fact that trend for Dev21-1 is fairly flat (Figure 6C) is often characteristic of a non-negligible detrital input [Wray and Wood 1998; Ducassou et al. 2019].

In the tectonic discrimination diagram (Figure 7A) presented by Schandl and Gorton [2002], the first group of previously defined (see above) samples plots in the “Within Plate Volcanic Zones” field while the second group plots in the “Active Continental Margin” field. In the tectonic diagram of Wood [1980], based on Hf/3–Th–Ta (Figure 7B), the second group of samples is characterized by a calc-alkaline chemistry, while the first group displays a higher Ta content relative to Th and Hf.

4.3. U–Pb dating of zircon

Unfortunately, it was not possible to find any minerals suitable for U–Pb dating in samples Per21-3 and Dev21-1.

Very few zircon grains were present in sample Per21-4 (Cinéritique Fm). Nine grains with various sizes were analyzed (Supplementary Table 3). They yielded variable U and Pb contents and fairly consistent Th/U ratios. With the exception of three discordant analyses (Figure 8A), three age groups can be defined: a first one around 2.0 Ga, a second one around 1.0 Ga and the last one around 0.6 Ga.

Again, very few zircon grains were found in sample Can21-5 (Cinéritique Fm). Thirteen grains with variable sizes were analyzed. They yield very high U (up to 3220 ppm) and Pb (up to 7648 ppm) contents (Supplementary Table 3) together with very high common Pb contents (f206c up to 39%). Their Th/U ratios are also highly variable (from 0.03 up to 0.53). Reported in a Concordia diagram (Figure 8B) they plot mostly in a discordant position with the exception of two grains (Zr12 and 13) which are concordant around 2.0 Ga.

Twenty-four analyses out of sixteen different zircon grains were acquired for sample Lat20-1 (Figure 3A; Supplementary Table 3). All zircon grains were prismatic, ranging in size between 70 and 200 μm and with simple concentric zoning (Figure 9). They yield variable U (236 to 9651 ppm) and Pb (50 to 908 ppm) contents with Th/U ratios consistent with a magmatic origin [0.1 to 0.26; see Witt et al. 2017 and references therein]. Eighteen analyses plot in a concordant position (Figure 8C) and yield a concordia date of 280.3 ± 2.6 Ma (MSWD = 1.2). Two analyses (Zr 1.1 and Zr 11.2, Supplementary Figure 1) were acquired on grains that contain a non-negligible common Pb content (f206c of 1.56 and 6.38% respectively) and that probably belong to the same age group. Two analyses acquired on the same grain (Zr2) plot in a concordant position with apparent ²⁰⁶Pb∕²³⁸U ages of around 320 Ma (Supplementary Figure 1), whereas a last analysis (Zr12, containing a non-negligible amount of common Pb; f206c% = 6.4), is highly discordant (52%; Supplementary Figure 1).

Thirty-one analyses (Supplementary Table 3) out of thirty-one different zircon crystals were acquired for sample Gal21-7 (Figures 1A and 2). Most crystals are rather elongated (up to 250 μm) and most of them display simple concentric magmatic zoning (Figure 9). Here again, the U (108 to 1886 ppm) and Pb (35 to 875 ppm) contents are highly variable. All the Th/U ratios, however, are compatible with a magmatic origin (0.13 to 0.48). A group of twenty-nine analyses (Figure 8D) define a concordia date of 278.4 ± 2.3 Ma (MSWD = 1.6). One grain (Zr5), characterized by a non-negligible common Pb content (f206c = 1.1%), plots to the right of the main batch of analyses and probably belongs to the same age group. The last analysis (Zr15) yields a concordant date around 470 Ma.

Thirty-two analyses (Supplementary Table 3) were acquired out of thirty-one zircon grains for sample Cam20-3 (Figure 3B). Zircon grains are generally prismatic with sizes ranging from 50 to 200 μm and display a magmatic zoning when imaged by cathode luminescence (Figure 9). They present variable U (212 to 1757 ppm) and Pb (41 to 1162 ppm) contents, with Th/U ratios ranging from “typical” metamorphic (0.03 to 0.09) to magmatic (0.16 to 0.62). Four age groups can be identified (Figure 8E and Supplementary Figure 1). The oldest zircon (Zr 27) yields an apparent ²⁰⁷Pb∕²⁰⁶Pb age of ca. 2646 Ma (Supplementary Figure 1). One grain (Zr 25) yields a concordant date of ca. 940 Ma (Supplementary Figure 1). A third group of six zircon grains (Zr 1, 7, 8, 11, 17 and 20) returns apparent ages around 450–460 Ma (Supplementary Figure 1), yielding the lowest (metamorphic?) Th/U ratios. The youngest group composed of twenty-three different zircon grains (Figure 8E) yields a lower intercept date of 278.7 ± 2.4 Ma (MSWD = 0.95).

Ten zircon grains were found in sample Gal21-6 (Figure 1A and 2; Supplementary Table 3). All the grains were prismatic with concentric magmatic zoning (not shown). They are rich in U (up to 2558 ppm) and Pb (up to 1393 ppm), with fairly consistent Th/U ratios characteristics of magmatic sources (0.21 to 0.66). With the exception of one grain (Zr1) with a concordant date around 340 Ma (Figure 8F), the remaining five concordant analyses yield a concordia date of 266 ± 7 Ma (MSWD = 0.7, n = 5), whereas the oldest concordant zircon from this group has a ²⁰⁶Pb∕²³⁸U apparent age of 277 ± 15 Ma. The four other analyses are discordant with variable amount of common Pb (f206c up to 2.78%).

4.4. Zircon Hf isotope composition

Three samples were selected for zircon Hf isotope measurements (Lat20-1, Cam20-3 and Gal21-7, Supplementary Table 3).

For each sample, ten zircon grains (eleven for sample Cam20-3), previously analysed for U–Pb, were targeted for Hf isotopes. Hf analyses were done as close as possible to the U–Pb ablation pits. The results are reported in Figure 10.

For sample Lat20-1, the analyzed zircon grains display very low ¹⁷⁶Lu∕¹⁷⁷Hf (0.0009 ± 1) and initial (calculated at 280 Ma) ¹⁷⁶Hf∕¹⁷⁷Hf ranging from 0.282600 to 0.282630, with nearly chondritic εHf(t) values (Figure 10) ranging from 0.8 to − 0.3 (mean 0.1 ± 0.2). T_DM model ages are broadly similar to those measured from the two other samples and display a mean value of 1269 ± 13 Ma.

Zircons from sample Gal21-7 (Supplementary Table 3) yield low ¹⁷⁶Lu∕¹⁷⁷Hf (0.0019 ± 3) and a mean ¹⁷⁶Hf∕¹⁷⁷Hfi of 0.282576 ± 12 (calculated at 278 Ma). εHf(t) are slightly sub-chondritic (Figure 10), ranging from − 0.3 to − 2.1 (mean = −1.2 ± 0.4). The T_DM mean value is 1352 ± 27 Ma.

For sample Cam20-3, the analyzed zircon grains present low ¹⁷⁶Lu∕¹⁷⁷Hf ratios (0.0024 ± 6) and homogeneous ¹⁷⁶Hf∕¹⁷⁷Hf_i ratios (0.282507 ± 15) calculated at 279 Ma. The corresponding initial εHf are subchondritic, with values ranging from − 4.8 to − 2.4, and a mean εHf_(t) of − 3.6 ± 0.5 (Figure 10). Their T_DM yield a mean value of 1507 ± 33 Ma. One xenocryst, dated at 454 Ma, was also analyzed and provides a distinct, slightly supra-chondritic εHf_(t) of 1.0.

5.1. Age and nature of the volcanism in the Saint-Affrique Basin

Two groups of samples can be distinguished among the studied samples.

The first group of volcaniclastic beds (Can21-5, Per21-3, Per21-4 and Dev21-1) belonging to the Cinéritique Fm (Figures 1A and 2), with a proposed Stephanian age (i.e. late Carboniferous), contains clasts (quartz and K-feldspars) that are consistent with a volcanic origin, and a fairly large detrital component. In the Zr∕TiO₂ versus Nb/Y diagram (Figure 5B, yellow ellipse) they plot in the field of trachyte/trachy-andesite, and they yield trace-element contents compatible with a felsic composition. They all have a calc-alkaline chemistry and exhibit “Within Plate Volcanic Zone”-like characteristics in the volcano-tectonic diagram of Wood [1980]; these features are compatible with a late to post-collisional volcanism. Sample Dev21-1, belonging to the Saint-Rome-de-Tarn Fm (i.e. proposed lower Cisuralian in age), exhibits a flat REE pattern when normalized to shale (Figure 6C), but also displays a “Within Plate Volcanic Zone” geochemical composition (Figure 7B). Unfortunately, it was not possible to date these samples because of the absence of volcanic zircon grains. The detrital zircon grains found in two of these samples demonstrates however the existence of a complex basement underneath and/or aside the Saint-Affrique Basin with ages ranging from the Paleo- to the Neo-Proterozoic (Figure 8A and B).

The second group of samples belong either to the Dourdou Fm (Lat20-1) or to the Saint-Pierre Fm (Gal21-6 and 21-7 and Cam20-3). These formations were so far attributed to the Sakmarian and Artinskian, respectively [Gand 1993]. Although these samples have a fair amount of muscovite indicating a terrigenous contribution, and can be sometime affected by a strong carbonation (Lat20-1), they all contain clasts of feldspars and quartz demonstrating the presence of a rather important volcanic content. These samples can be classified as dacites/rhyodacites, have a strong negative Eu/Eu* anomaly, a significant LREE enrichment and a fairly flat HREE spectrum; these features are all in a good agreement with their felsic nature, and can be defined as calc-alkaline rocks. All of these features are in a good agreement with a post-orogenic deposition setting. All of these samples provided volcanic zircon grains. The stratigraphically oldest sample from this group (Lat20-1) records a volcanic episode during the Kungurian. The obtained date of 280 ± 2.6 Ma for this sample is interpreted as the age of the deposition of the bed, assuming that the dated zircon grains were not reworked after the volcanic eruption [e.g Rossignol et al. 2019]. Applying the same assumption, the samples at the top of the sedimentary succession were then deposited at 278.4 ± 2.3 Ma (Gal21-7) and 278.7 ± 2.4 Ma (Cam20-3) respectively, i.e. also during the Kungurian. The date of 266 ± 7 Ma found for sample Gal21-6 may be related to slight Pb losses enhanced by the zircon’s high U content (>1500 ppm) and has to be considered as a minimum age of deposition, as this sample belongs to the same ash bed as Cam20-3. Therefore, we propose that the date of 277 ± 15 Ma found for the oldest concordant zircon from this sample is the best estimate for the deposition age of this volcanic ash bed.

The mean εHf_(t) calculated for the zircon grains from the three dated volcanic ash beds Lat20-1, Gal21-7 and Cam20-3 provides chondritic to sub-chondritic values of 0.1 ± 0.2; − 1.2 ± 0.4 and − 3.6 ± 0.5, respectively. The corresponding Hf model ages of 1.3–1.5 Ga represent an average of the sources involved in magma genesis or assimilated during the travel path upward to the surface of the magmas. In any case, this indicates a significant degree of recycling of Proterozoic or older material, which is also consistent with the occurrence (in the three volcanic levels) of zircon grains with apparent ages ranging from the Archean to the Neoproterozoic. It is however difficult to evaluate whether the Hf isotope signature of the analysed zircon grains reflects either a source signature or an open system evolution, such as assimilation of an older crust by juvenile, mantle-derived, magmas. Clarification of this point is key to the interpretation of the geochemical data, as it could indicate crustal reworking processes, or new crustal growth by addition of material issued from the mantle during the post-collisional Variscan period. Although sampled at different stratigraphical levels, samples Gal21-7 and Cam20-3 belong to the same F4/F5 formation and display remarkably similar U–Pb ages (278.4 and 278.7 Ma, respectively) and trace- and major-element chemistry. Both samples display subchondritic, but substantially different, Hf isotope signatures (εHft = −1.2 ± 0.4 and − 3.6 ± 0.5 respectively). This difference indicates that magma genesis involved partial melting of a compositionally heterogeneous crustal section. The mean age for this crustal section is estimated between 1.3–1.6 Ga, assuming felsic precursors (with a ¹⁷⁶Lu∕¹⁷⁷Hf ratio of 0.015), or 1.6–2.0 Ga, assuming mafic precursors (with a ¹⁷⁶Lu∕¹⁷⁷Hf ratio of 0.021). Sample Lat20-1 yields a chondritic value that can be explained in two ways. The first is a mixing (at c. 280 Ma) between a suprachondritic mantle-derived magma and subchondritic Paleoproterozoic or older crustal sources in proportions that coincidentally produced a εHf chondritic value. An alternative possibility is a remelting (at c. 280 Ma) of a 1.3 Ga old or 1.6 Ga old Paleoproterozoic source, which would have evolved with a calculated ¹⁷⁶Lu∕¹⁷⁷Hf ratio of, respectively, 0.015 assuming a felsic precursor, or a ratio of 0.021, typical of the time-integrated evolution of a mafic (lower) crust (Figure 9). It is worth noting that the vertical scattering observed for all analyses in Figure 9 (from 0.8 to − 4.8) may reflect such a mixing trend between mantle- and crustally-derived components. Whichever hypothesis is retained, it requires a heat source that can be found in mantle-derived magmas present at the base of the continental crust, as mentioned in many previously published studies [e.g. Breitkreuz and Kennedy 1999]. Therefore, although our results indicate intra-crustal remelting and recycling of an old Paleoproterozoic to Archean basement, the addition of juvenile magmas triggering partial melting of the lower crust or mixed with crustally-derived magmas cannot be ruled out.

Although this needs further investigation, the Hf analyses tend to indicate an increase of crustal input in the production of the magmas, with εHf chondritic values for the oldest sample (Lat20-1) becoming increasingly negative towards the top of the basin. Interestingly, there is also a correlation between εHf(t) and Eu/Eu*, which may indicate that the magmas producing the stratigraphically most elevated volcanic ash beds Gal21-7 and Cam20-3 underwent fractional crystallization processes involving plagioclase, more pronounced than the other samples, and that they possibly had more time to interact with, and assimilate crustal material.

The first group of samples consists of trachy-andesitic lavas, which are characterized by a greater HREE depletion and smaller Eu/Eu* anomalies than the rhyodacitic samples of group 2 (including the dated samples Lat20-1, Gal21-7 and Cam21-3). The greater HREE depletion of group 1 samples may indicate that the magmas were generated at a greater depth than samples from group 2. Lastly, the identification of Neoproterozoic to Archean zircon grains in the volcanic ashes substantiates the occurrence of a complex and old basement both underneath and alongside the basin.

The different geochemical signatures of the two groups mark a significant change in the petrological processes producing the magmas, especially an evolution of the main magmatic sources involved, as the geodynamical conditions evolved from a late- to a post-orogenic setting with time. The “Within Plate”- versus “Active Margin”-like signatures, coupled with our Hf data interpretations in the younger group, reflect that the crustal contribution increased as time proceeded. In any late- to post-orogenic setting, this has to be controlled by substantial modifications of the geothermal gradient, and a significant increase of the heat flux, which can hardly be conceived without a simultaneous increase of the production of mantle melts triggering crustal melting. Because the production of mantle magmas at high rates may induce significant crustal melting, it is likely that from the base to the top of the sedimentary succession, the Saint-Affrique pyroclastic layers recorded the transition from a late-orogenic extension to a post-orogenic extension at much higher rates as the basin enlarged.

Our new constraints on the age and nature of the volcanism recorded in the sedimentary succession also questions the location of the emission centres of the volcanic ashes, in the absence of known volcanic edifices near the Saint-Affrique Basin, feature which is a recurrent issue in Western Europe Carboniferous and Permian continental basins [e.g. Koniger and Stollhofen 2001; Pellenard et al. 2017]. Explosive subaerial calc-alkaline volcanic activity with Permian ages is known in the Central Pyrenees (about 200 km from the Saint-Affrique Basin), where it is associated with high-intensity explosive magmatic eruptions of rhyodacitic to rhyolitic lavas and associated widespread pyroclastic flows and overriding ash clouds [Marti 1996; Pereira et al. 2014], and transported several hundreds of kilometres under the dominant wind. On the other side, Permian subaerial calc-alkaline volcanic episodes, including andesitic, dacitic and rhyolitic ignimbrite lavas are known in northwestern Corsica and Sardinia [Cabanis et al. 1990; Timmerman 2004, and references therein], and in the Esterel Massif [Zheng et al. 1992; Nmila 1995], and could also represent good candidates for the volcanic ash input. Nonetheless, the possible pyroclastic flow nature of the sampled layer from the Dourdou Fm (sample Lat20-1) suggests that edifices that were more proximal to the basin still have to be identified.

5.2. Implications for the deposition of the Saint-Affrique Basin

The new radiochronological ages presented in this study refine the chronological setting of the Saint-Affrique Basin (Figure 11), the sedimentary successions of which were considered as deposited from the early to the middle Permian [Rolando et al. 1988, Figures 2 and 11], and then during the early Permian only [Gand 1993, Figures 2 and 11]. Our new ages encompass the Kungurian Stage (uppermost Cisuralian, Figure 11). The oldest age (Lat20-1, 280.3 ± 2.6 Ma), i.e. in the lower part of the Dourdou Fm, is, within errors, very close to the lower limit of the Kungurian, given that the Artinskian–Kungurian boundary is set at 283.5 ± 0.6 Ma [Cohen et al. 2013, updated; Figure 11]. No age could be determined at the base of the sedimentary succession, but it is likely that Asselian to Sakmarian (lower Cisuralian) deposits are preserved.

The top of the sedimentary succession of the Saint-Affrique Basin is not dated, but considering the youngest age found in this study (Cam20-3, 278.7 ± 2.4 Ma), located less than 100 m below the youngest preserved deposits, as well as the absence of significant erosional surfaces or discontinuity in the sedimentation above this volcanic level, it is highly probable that the top of the sedimentary succession was also deposited during the Kungurian, the upper limit of which is set at 273.01 ± 0.14 Ma [Cohen et al. 2013, updated].

The maximum duration of the sedimentary deposition between the stratigraphically lowest and highest dated volcanic ash beds levels, i.e. Lat20-1 (280.3 ± 2.6 Ma) and Cam20-3 (278.7 ± 2.4 Ma), is 6.6 Ma. During this timeframe, the sediment thickness estimated by coupling borehole data and spatial correlations [based on the 1/50,000 geological map of Guérangé-Lozes et al. 1995] is about 1100 to 1500 m. Considering this maximum duration, the sedimentation rates (compacted sedimentary succession), appear to be low, ranging from 0.160 mm/yr to 0.230 mm/yr. It is worth noting the mean sedimentation rate estimated for other late-Carboniferous–early Permian basins presenting a similar geological setting: in the northeastern Massif Central Autun Basin [Mercuzot 2020], sedimentation rates based on precise radiometric ages were estimated at 0.45 mm/yr, taking the compacted sedimentary succession into account, and at 1.1 mm/yr when estimating a decompacted succession thickness. In the Saar-Nahe Basin (western Germany) the sedimentation rate is also higher, i.e. 0.3 mm/yr (compacted sedimentary succession). Some modern lacustrine systems, e.g. the Lake Titicaca (Altiplano, late-orogenic setting) and the North-American Great Lakes (Superior, Erie and Huron, glacial lakes) exhibit sedimentation rates between 0.1 and 7.4 mm/yr [Kemp and Harper 1976; Kemp et al. 1977, 1978; Durham and Joshi 1980; Lojka et al. 2009] that are consistent with our calculated rates. It also demonstrates that the sedimentation rates calculated in this study might be underestimated because of the lack of precise petrophysical and burial condition data. Consequently, the discontinuity between F3 and F4 fms (Figures 2 and 11) should not represent a substantial hiatus.

These new chronostratigraphical data clearly indicate the lack of the Middle (Guadalupian) and Upper (Lopingian) Permian deposits in the present-day sedimentary succession. Moreover, if we consider a middle-Anisian age for the basal Triassic sandstone deposits, as in the Lodève Basin [Lopez and Mader 1985], the angular unconformity between the youngest Permian and the Triassic deposits represents a major hiatus of about 23 Myrs, during which the residual Variscan reliefs and the upper part of the Permian sedimentary succession of the Lodève and Saint-Affrique basins were eroded and peneplaned. In the Lodève Basin, the overall thickness of the eroded sedimentary succession between the top of the Permian deposits and the middle-Anisian unconformity is estimated at 1000 to 1500 m [Lopez et al. 2008, Figure 2].

In Europe, increasing correlations are tentatively made between Carboniferous–Permian basins, based on both radiometric ages and biostratigraphic data [e.g. Schneider and Scholze 2018; Pellenard et al. 2017; Ducassou et al. 2019; Schneider et al. 2020]. The Lodève Basin is one of the most used basin to establish correlations in Western Europe, mostly because of its almost complete sedimentary succession from the late Carboniferous to the late Guadalupian [i.e. late middle Permian; Schneider et al. 2020]. However, the present re-evaluation of the ages of the Saint-Affrique Basin, located in the direct vicinity and formed in the same geodynamic setting, calls for new radiometric age constraints in the Lodève Basin.