Discovery of Variscan orogenic peridotites in the Pelvoux Massif (Western Alps, France)

– Small bodies of mantle-derived peridotites and other ultrama ﬁ c rocks are commonly found in exhumed lower crustal units of collisional orogens. They provide a direct record of the complex evolution of the upper mantle before and during an orogeny, and are therefore key markers of the geodynamic evolution of an orogen. We report here the discovery of such mantle-derived peridotites, which occur as fragmented enclaves in migmatites of the high-grade Variscan lower crust exposed in the Pelvoux Massif (external Western Alps). A wide petrographic diversity has been observed, from very fertile, garnet-bearing lherzolites, to more depleted spinel/chromite-bearing harzburgites. Thermobarometric calculations on a garnet lherzolite indicate an initial stage at 3.0 – 4.0 GPa and 970 – 1140 ° C, followed by exhumation to 0.8 – 1.3 GPa and 800 – 850 ° C, while the harzburgites do not show any evidence of equilibration in the garnet ﬁ eld. Petrological observations, whole-rock geochemistry and in situ mineral compositions suggest the peridotites have undergone a complex history prior to their incorporation in the lower crust during the Variscan Orogeny. They derive from a refractory mantle, which has experienced variable degrees of melt depletion, and has then been extensively refertilized. Cryptic metasomatism is observed in all samples. It is characterized by an enrichment in large-ion lithophile elements (LILE, in particular Cs, Rb, U and Pb) relative to high ﬁ eld strength elements (HFSE), in particular Nb and Ta. This cryptic metasomatism is presumably related to percolation of subduction-related ﬂ uids or melts in the mantle. In addition, modal metasomatism occurred in some samples, where crystallization of phlogopite, pargasite, chromite and apatite has been observed. This modal metasomatism resulted in signi ﬁ cant enrichment in K 2 O, P 2 O 5 and Cr 2 O 3 of the bulk rock, together with a strong enrichment in incompatible LREE relative to HREE. These geochemical characteristics are strikingly similar to that of syn-collisional, Mg – Cr – LILE rich mantle-derived (ultra)-potassic magmas such as durbachites and vaugnerites, which are ubiquitous in the Variscan metamorphic allochthons of Massif Central, external Alps, Vosges and Bohemian Massif. We therefore suggest that this metasomatism results from dynamic percolation of the peridotites by K 2 O – P 2 O 5 – Cr 2 O 3 - rich melts from which the durbachites and vaugnerites are primarily derived. These geochemical characteristics are in line with whole-rock Nd isotopic compositions, which indicate enrichment of the mantle by a continental crust component, presumably related to Variscan subductions. This evolution is consistent with that of other Variscan peridotites in the Eastern Alps (Ulten) and the Bohemian Massif, where multiple metasomatic episodes related to melts or ﬂ uids released in Variscan subduction zones have been documented.


Introduction
Occurrence of mantle-derived ultramafic slices, lenses or boudins is a common feature of high-grade lower crustal units exposed in collisional orogens worldwide (e.g.Brueckner and Medaris, 2000;Nimis and Trommsdorff, 2001;Medaris et al., 2005;Chen et al., 2015;Brueckner, 2018).Although they only represent a tiny fraction of these high-grade units, ultramafic bodies provide extremely valuable information about metasomatic and magmatic processes occurring in the upper mantle during and before an orogeny (Malaspina et al., 2006;Scambelluri et al., 2006;Sapienza et al., 2009;Brueckner et al., 2010;Kube s et al., 2022).They may also retain pressuretemperature information and thus provide constraints about burial and exhumation of (ultra)-high-pressure (U-HP) units in collision zones.They are therefore crucial markers to constrain the geodynamic evolution of an orogen.
In the Variscan Belt (Fig. 1a), small ultramafic mantle bodies, including garnet peridotites, spinel peridotites and pyroxenites, are found in the metamorphic allochthonous units that form exhumed portions of the lower-to-mid-crustal orogenic root, mainly in the Bohemian Massif (Medaris et al., 2005(Medaris et al., , 2015;;Kube s et al., 2022), the Vosges-Black Forest Massif (Altherr and Kalt, 1996;Altherr, 2021) and the French Massif Central (Gardien et al., 1990;Godard, 1990).In detail, these mantle bodies are found in exhumed lower to midcrustal units from different tectonic domains, which record different metamorphic evolutions (Medaris et al., 2005;Lardeaux et al., 2014;Altherr and Soder, 2018).In the Bohemian Massif, where the largest number of these ultramafic bodies has been found, a distinction has been recognized between peridotite-bearing units exhumed close to the plate interface (Saxo-Thuringian suture) and those exhumed further south, between the former Devonian subduction arc and the Brunia continental backstop (Moldanubian zone, Schulmann et al., 2009).The former typically record colder peak-P conditions (< 800 °C) and shorter-lived peak-T than the latter, which record peak-P at UHT conditions (850-1050 °C), followed by near-isothermal decompression (Maierová et al., 2021, and references therein).Therefore, these different units correspond to contrasted exhumation mechanisms, and mark different processes within the Variscan collision.While the plate-interface probably correspond to a former subduction mélange exhumed in the subduction channel, exhumation of Moldanubian peridotites and other UHP rocks might involve less conventional mechanisms of lower crustal "relamination" (Schulmann, Lexa et al., 2014;Maierová et al., 2018Maierová et al., , 2021)).
Multiple studies of the mantle-derived fragments have been undertaken in the Bohemian Massif (e.g.Medaris et al., 2005Medaris et al., , 2015;;Ackerman et al., 2020;Kube s et al., 2022), which may therefore be a good reference site to discuss these processes.In other Variscan Massifs, deep crustal and mantle processes have been comparatively less studied, mainly because of the scarcity of well-preserved mantle bodies.This is particularly the case in the southeastern Variscides, where there is virtually no study on mantle fragments, apart from the garnet peridotites of the Ulten Zone, in the austroalpine basement (Godard et al., 1996;Tumiati et al., 2003;Scambelluri et al., 2006).We document here the discovery of new outcrops of peridotites in the Variscan basement of the external Western Alps, in exhumed lower crustal units of the Pelvoux massif (Fig. 1b).These peridotites occur as fragmented enclaves in migmatites, and some of them have been moderately affected by serpentinization.This contribution aims to provide field, petrological, thermobarometric, and whole-rock geochemical and isotopic data of these enclaves, in order to discuss their metamorphic and metasomatic evolution.This evolution will then be compared with that of other Variscan mantle bodies, in particular from the Bohemian Massif, in order to provide some insights into the evolution of the Variscan mantle in the external Western Alps.Guillot and Ménot (2009).The investigated area (black rectangle) is located in the central part of Pelvoux Massif (Fig. 2), but two samples of serpentinites were collected in the southern part of Belledonne.Guillot et Ménot (2009).La zone étudiée (rectangle noir) est située dans la partie centrale du massif de Pelvoux (Fig. 2), mais deux échantillons de serpentinites ont été collectés dans la partie sud de Belledonne.

General setting
The Pelvoux Massif is a well-exposed portion of Variscan middle-to-lower crust located in the French Western Alps (Fig. 1b).It belongs to the External Crystalline Massifs (ECM), which represent the exposed Paleozoic basement in the external Western Alps.The Alpine overprint in these massifs is generally considered to be very mild, and consists mostly in localized brittle/ductile deformation, which occurred at greenschist-facies conditions (< 350 °C, Bellahsen et al., 2014;Bellanger et al., 2015).The Paleozoic basement of the ECM has been strongly overprinted during the Variscan orogeny (Guillot and Ménot, 2009;Jacob et al., 2022), and displays similarities with the metamorphic allochthonous units exposed in the French Massif Central, Vosges and Bohemian Massifs (Faure et al., 2009;Skrzypek et al., 2012;Lardeaux et al., 2014;Vanderhaeghe et al., 2020;Martínez Catalán et al., 2021).This includes in particular: (i) the nature of protoliths, which mainly consist of Cambrian-Ordovician magmatic and sedimentary sequences with widespread tholeitic magmatism, marking a stage of rifting along the northern margin of Gondwana (Von Raumer and Stampfli, 2008); (ii) the presence of ophiolitic relics, the largest one being the ophiolite complex of Chamrousse (496 ± 6 Ma) in the massif of Belledonne (Pin and Carme, 1987;Ménot et al., 1988), which lies on the top of the metamorphic nappe pile and is devoid of HP metamorphism (Fig. 1b); (iii) the widespread occurrence of (retrogressed) eclogite/ mafic HP granulite bodies in the high-grade lower crustal units.These eclogites record a Visean (ca.340-330 Ma) HP stage with peak-P conditions estimated around 1.4-1.8GPa and 650-750 °C (Rubatto et al., 2010;Jouffray et al., 2020;Jacob et al., 2021Jacob et al., , 2022;;Vanardois et al., 2022).This eclogitisation is interpreted to mark the closure of a shortlived marginal basin opened in a back-arc setting on the Gondwanian plate at ca. 350 Ma (Fréville et al., 2018), which would have been squeezed and incorporated into the Variscan orogenic wedge by southeastward propagation of deformation during the lower to middle .Felsic protoliths do however not record HP conditions.In weakly to non-migmatized gneiss and micaschist of the Western Pelvoux and Belledonne Massif, a Visean (ca.340-330 Ma) MP-MT stage (0.5-1.0 GPa and 550-680 °C) is recorded, with no relics of HP/UHP assemblages (Guillot and Ménot, 1999;Fréville et al., 2018Fréville et al., , 2022)).In the more highly migmatized units of inner Pelvoux, this MP-MT stage is overprinted by a HT granulitic stage at ca. 800-850 °C and 0.6-0.9GPa (Grandjean et al., 1996;Jacob et al., 2022).
As in the rest of the Variscan metamorphic allochthons, Devonian-Carboniferous magmatism is widespread in the ECM.First, the opening of Devonian-Tournaisian basins, presumably in a back-arc setting, was associated with intrusion of trondhjemite dykes and sills around 350 Ma (Fréville et al., 2018).Tonalite plutons of the same age have also been dated in the Aar Massif (Ruiz et al., 2022).This low-K magmatism was followed by emplacement of K-rich sub-alkaline granitoids from the Visean to the early Permian (ca. 345-295 Ma, Debon and Lemmet, 1999).A large part of this magmatism results from crustal anatexis during the syn-to post-collisional stages, as evidenced by the large proportion of migmatized lower crust exposed in the ECM (Guillot and Ménot, 2009).However, a significant contribution of mantle sources is also evidenced by the presence of (ultra)-potassic (UHK), mafic to intermediate igneous rocks, locally referred to as durbachites or vaugnerites (Debon et al., 1998;Von Raumer et al., 2014).These rocks are particularly rich both in compatible (Mg, Ni, Cr) and incompatible elements including large-ion lithophile elements (LILE: Cs, Rb, Ba, K), Th, U and light rare earth elements (LREE: La, Ce, Nd, Sm).Thus, the most widely accepted view attributes this magmatism to the melting of an enriched, metasomatized mantle source contaminated by subducted crustal material (Janou sek and Holub, 2007;Soder and Romer, 2018).

The Pelvoux Massif and peridotite-bearing localities
The Pelvoux Massif lies in the eastern part of ECM and exposes high-grade metamorphic series composed of highly migmatized mafic to felsic protoliths (metasediments, orthogneiss, amphibolites), intruded by several granitoid plutons (Fig. 2).The metamorphic series display different sets of N-N150°subvertical and subhorizontal migmatitic foliations, which are consistent with longitudinal flow of partially molten material in a N-N150°sinistral transpressive setting (Fréville et al., 2022;Jacob et al., 2022).However, the early synmigmatitic fabrics have been partly overprinted by later, subsolidus deformation occurring in the same sinistral wrenching setting, and by alpine ductile/brittle tectonics (Fig. 2).Occurrence of ultramafic rocks, mainly serpentinites, has been acknowledged for long in the Pelvoux Massif (Pecher, 1970;Le Fort, 1971), as well as in surrounding massifs of Belledonne and Aiguilles Rouges (Barfety et al., 2000;Von Raumer and Bussy, 2004).Inspection of screes in multiple areas indicates that these serpentinites are ubiquitous in the metamorphic basement of the Pelvoux Massif, although in very subordinate volume (Fig. 2).When observed in place, they mostly occur as meter to decameter-wide lenses, which can be up to a hundred meters long.Some of them are concordant with the surrounding metamorphic foliation, but they are commonly transposed in brittle fault zones, as they acted as a weak layer localizing deformation (Fig. 3e).Thus, most of these serpentinites have been strongly deformed and retrogressed, and do not preserve any relics of olivine, pyroxene or any other primary mineral that existed before serpentinization.A second type of peculiar ultramafic rocks was reported by Le Fort (1971) and Pecher (1970) as "amphibolite boulders with a concentric structure".They occur as decimeter-size dark boulders in migmatites, and consist of a core composed of amphiboles (actinolite-tremolite, hornblende, anthophyllite), chlorite, iron oxides, phlogopite, calcite, and a few relics of spinel and augite, which is surrounded by centimeter-wide reaction coronas of tremolite and biotite.The rocks described by Le Fort (1971) were obviously heavily retrogressed and did not preserve much of their initial assemblages.Thus, their origin remained unclear.They were regarded either as relics of metasomatized calcsilicate boudins, or as ultramafic cumulates of gabbro-dolerite series similar to the Chamrousse ophiolite in the massif of Belledonne.Extensive investigation of the screes in the area covered by the map in Figure 2 has led us to find well-preserved specimens of these boulders, with limited reactions with the migmatites, which contain a primary assemblages dominated by olivine, orthopyroxene and clinopyroxene, with spinel and/or garnet.It is therefore clear that these rocks are retrogressed peridotites.The main occurrences have been found in screes on the southern slopes of Valgaudemar Valley, between the summits of Olan and Les Rouies, and in the Bans Valley.Two sites where they are in place are shown with yellow stars in Figure 2. The first one is located on the eastern side of the Sellar pass, in the Bans Valley (GPS: N44.83612°, E6.33910°, alt.2770 m), and the second one above Refuge de l'Olan (GPS: N44.85089°, E6.20314°, alt.2744 m).These peridotites form rounded to sub-angular 5 cm to 1 m wide enclaves.They generally occur in highly-mobilized part of migmatites (diatexites), and commonly appear as "nest" of enclaves, without any particular orientation (Figs.3a-3c).This disposition suggests the enclaves represent pieces of larger peridotite bodies, possibly up to decameter-size, which have been fractured during injection of granitic magmas and dismembered into smaller blocks.The enclaves are neither particularly elongated (aspect ratios between 1 and 1.8) nor oriented in any preferential direction, which suggests they did not experience significant ductile deformation in the migmatite.Reaction with the hosting migmatite is evidenced by the development of coronas around the peridotite enclaves.However, they are not systematic.Some enclaves are almost devoid of these coronas, and display instead sharp contacts with the migmatite (Fig. 3b), while hornblende ± biotite reaction coronas may develop in other cases (Figs.3a and 3c).The most spectacular ones have been observed on peridotites from the Sellar pass, and consist of a 1-3 cm wide coronas of radially-oriented tremolite, surrounded by a ca.0.5-1 cm concentric layer of biotite (Figs.3c and 3d).However, these coronitic reactions only affected the enclaves at their outer rims or along fracturation pathways, leaving the inner parts of largest enclaves (> 10 cm) relatively well-preserved (Figs.3f and 3g).

Samples and methods
Eleven samples of peridotite and 4 samples of serpentinite were collected for petrological study and whole-rock geochemical and isotopic analyses.All the samples come from the Pelvoux Massif, excepted two samples of serpentinite collected in a large hectometer-size lense in the Belledonne Massif.Sampling coordinates and petrological classification of each sample are summarized in Table 1.

Petrography and mineral analysis
Backscattered electron (BSE) images were acquired at ISTerre Grenoble with a Tescan Vega 3 scanning electron microscope equipped with an energy dispersive spectrometer (EDS) for semi-quantitative analysis.It was operated with an accelerating voltage of 16 kV and a beam current of 10 nA.Inclusions in garnet were identified by Raman spectroscopy, using a Horiba Labram Soleil instrument installed at ISTerre Grenoble, with a 532 nm laser beam.Mineral compositions were measured using a JEOL JXA-8230 electron probe microanalyzer (EPMA) at ISTerre Grenoble, which is equipped with five wavelength dispersive spectrometers (WDS) and an additional EDS.All silicates except olivine were analyzed using standard EPMA protocols, with an accelerating voltage of 15 kV (amphibole, phlogopite) to 20 kV (garnet, pyroxenes), a beam current of 10 nA (amphibole, phlogopite) to 20 nA (garnet, pyroxenes), and a counting time per element of 30 to 60 s for peak and both background positions.Natural minerals and synthetic glass were used for standardization, and the ZAF matrix procedure was applied for data reduction.The spot size was set to 1-3 mm depending on the size of minerals and the presence of volatile elements.The composition of olivine was analyzed following highprecision method for major (Si, Mg, Fe), minor (Ni, Mn) and trace (Al, Ca, Cr, Ti, P) elements (Batanova et al., 2015) at an acceleration voltage of 20 kV, beam current of 900 nA and beam diameter of 2 microns.Total analysis time of each point was 7.2 minutes.MongOl sh11-2 olivine reference material (Batanova et al., 2019) was used as primary standard to verify precision and accuracy (e.g.Sobolev et al., 2007).Deviation from the reference measured during the analytical session was between 2 and 13 ppm for trace elements, between 10 and 85 ppm for minor elements and between 0.01 and 0.06 wt% for major oxides, which is in any cases below the 2-sigma uncertainty reported in Batanova et al. (2019).The composition of Cr-spinel was measured at 20 kV and 50 nA beam current.Counting time was adapted to element content and was between 60 to 120 s for each element (for peak and both background positions).The standardization was made using a synthetic oxide standard set (P&H Developments Ltd., Calibration Standards for Electron Probe Microanalysis, Standard Block GEO) for all elements except Mn (on rhodonite).Ferric iron in spinel was calculated assuming perfect stoichiometry.Repeated measurements of the chromite USNM 117075 (Jarosewich et al., 1980) standard and spinel samples Bar 8601-10 and Dar 8502-2, whose Fe 3þ /Fe total ratio had been measured by Mossbauer spectroscopy (Ionov and Wood, 1992) have shown that the selected method provides Fe 3þ /Fe total ratios accurate to within the measurement error.Spinel Bar 8601-10 (Ionov and Wood, 1992) was used as a primary standard and was analyzed multiple times during the session to verify precision and accuracy.Deviation from the reference value measured during the session was below 50 ppm for most of trace elements (Ti, V, Mn, Ni) and bit higher for Zn (66 ppm) and Si (86 ppm).For major oxides, deviation from the reference value was about 0.64 wt% for Al 2 O 3 and 0.06 wt% for FeO and MgO.

Whole-rock composition
Samples were cut to remove the coronas formed at the contact with the migmatites, and domains devoid of large veins were carefully selected to study only the best-preserved domains.They were then crushed in a jaw crusher and pulverized into a 70-80 mm powder using an agate crusher.
whole-rock geochemical analyses were performed at the SARM in Nancy, using a Thermo Fischer iCap6500 ICP-OES for major oxides and an iCapQ ICP-MS for minor and trace elements.Details about the analytical procedures for wholerock geochemical analyses are described in Carignan et al. (2001).Samples display high loss on ignition between 7 and 13% due to serpentinization, and therefore the analyses were recalculated on an anhydrous basis to allow comparison between samples.Raw analyses are provided in the Supplementary Material.The R-based GCDKit software (Janou sek et al., 2016) was used for statistical analysis and plotting of whole-rock geochemical data.

Sr-Nd isotopes
Nine samples were selected for Sr-Nd isotope analyses.Due to low Nd concentration, only five of them were analyzed for Nd isotopes.Isotopic measurements were also carried out at the SARM in Nancy.100 to 200 mg of powdered samples were dissolved into a mixture of concentrated HNO 3 and HF heated to 115 °C during 24 to 48 h, followed by concentrated HCl at 125 °C during 24 h.
Sr and Nd were isolated via ion-exchange chromatography using Sr.Spec, Tru.Spec and Ln.Spec resins, following the procedure described in Pin et al. (1994) and Pin and Zalduegui (1997).Sr and Nd isotopes were analyzed by thermal ionization mass spectrometry (TIMS), using respectively a Triton Plus and a Neptune Plus instrument operated in static multi-collection mode.The 87 Sr/ 86 Sr ratios were corrected for mass bias assuming 87 Sr/ 88 Sr = 0.119400, and 143 Nd/ 144 Nd ratios were corrected to 146 Nd/ 144 Nd = 0.721900 (Luais et al., 1997).Internal standards NBS 987 and JNdi-1 (Tanaka et al., 2000) were used for Sr and Nd, respectively.The decay constants applied to age-correct the isotopic ratios are from Steiger and Jäger (1977) for Sr and Lugmair and Marti (1978) for Nd.The eNd i values were obtained using composition of the chondritic uniform reservoir (CHUR) of Bouvier et al. (2008).

Petrography
All the samples collected in the ultramafic enclaves have been retrogressed to various degrees and contain a high amount of volatiles (loss on ignition between 7 and 13 wt%), which reflects fluid addition during retrogression.Typical serpentinization textures are observed, with the development of a mesh of serpentine þ Fe-oxide in cracks and at olivine grain boundaries.Minor amount of carbonates are also observed in the most retrogressed samples.Four different types of ultramafic rocks have been identified based on petrological observations: serpentinites (4 samples), retrogressed garnet lherzolite (9 samples), retrogressed spinel harzburgite (1 sample) and retrogressed phlogopite-chromite harzburgite (1 sample; Fig. 4).The serpentinites present whole-rock major element compositions similar to the harzburgites (see Sect. 6, Geochemistry Section below), but rarely preserve relics of their primary minerals.They are mostly composed of HT serpentine (antigorite) and Fe-oxides, with a few relics of spinel.Petrological descriptions are therefore focused on the peridotite samples.

The garnet lherzolites
The garnet lherzolites are the most distinctive type in the field.Macroscopically, they are composed of a dark green, fine-grained matrix containing millimeter to centimeter-size reddish spots (former garnet) and millimeter-size grains of  diopside (Figs.3f and 4a).The dark matrix contains 100-200 mm grains of enstatite (labeled Opx-I in Fig. 5a), diospide (Cpx-I) and olivine (Fig. 5a).Garnet is almost completely destabilized, and only a few millimeter-size relics were identified (Fig. 4).Most of the former grains (reddish spots) have been replaced by kelyphites mainly composed of micrometer-size intergrowth of spinel and enstatite (Opx-II in Fig. 5b) with subordinate amounts of larger grains (10-20 mm) of diopside (Cpx-II) and amphibole.These kelyphites are surrounded by a coarse-grained corona of enstatite (Opx-II) þ diopside (Cpx-II) (Figs.5a and 5c).In most samples, the fine-grained kelyphite has partially recrystallized into a coarsegrained globular symplectite of orthopyroxene þ spinel ± clinopyroxene ± amphibole and eventually evolved toward a coarse-grained polygonal assemblage of the same minerals with relics of vermicular spinel (Figs.4b and 5c).The rare garnet relics are directly surrounded by a second generation of symplectite composed of plagioclase þ amphibole þ Cr-rich spinel (Fig. 5b).The garnet relics contain small inclusions (ca.20 mm wide) of orthopyroxene, which themselves contain tiny dark inclusions of graphite, identified by Raman spectroscopy (Fig. 5d).

The spinel-pargasite-(phlogopite) harzburgite (JB-19-22)
The harzburgite sample JB-19-22 is mostly composed of olivine, enstatite, pargasitic amphibole and Mg-spinel, with rare diopside, phlogopite, and accessory Fe-oxides (Figs. 4a, 6a and 6b).Spinel forms 200-500 mm grains that contain inclusions of pargasitic amphibole (Fig. 6b), but never occurs in symplectites or kelyphites like in the garnet lherzolites, which suggests that garnet was absent from the primary assemblage.Olivine forms 50-200 mm grains, which have been partly serpentinized (Fig. 6a).Enstatite is abundant and forms large grains up to 500 mm wide.The largest grains are preferentially localized in discontinuous, millimeter-wide bands (Fig. 4a).By contrast, diopside is extremely rare and only a few grains < 100 mm, rimmed by amphibole, were identified (Fig. 6c).Pargasite forms 100-300 mm grains and occurs as smaller inclusions in spinel.Phlogopite is scarce and generally found in association with pargasite.

The phlogopite-chromite harzburgite (JB-19-21)
The phlogopite-chromite harzburgite contains olivine, enstatite, phlogopite, chromian spinel (chromite) and accessory apatite, and is completely devoid of clinopyroxene (Figs.4a  and 6d).The sample investigated in detail (JB-19-21) is an enclave surrounded by a thick corona of biotite and tremolite (Figs.3c, 3d and 4a), but large grains of phlogopite are also visible in some enclaves devoid of large reaction coronas (Fig. 3g).Olivine forms partly serpentinized grains 200-500 mm in size (Fig. 5d).Enstatite and phlogopite form large grains up to 1 mm-wide.Chromite is ubiquitous.It forms subeuhedral grains ranging from 10 mm to > 100 mm, which occur either in the serpentinized matrix or as inclusions in phlogopite, and contain itself small inclusions of phlogopite (Fig. 6e).Apatite is also closely associated with phlogopite, and generally occurs in contact with this mineral (Fig. 6d).Harzburgitic domains are cut by centimeter-wide sub-parallel veins of orthopyroxenite (Fig. 3a), which are mainly composed of large grains of enstatite (up to 3 mm wide), with subordinate olivine, phlogopite and chromite (Fig. 6f).Partial replacement of Mg-rich primary olivine by, BSE-brighter Fe-rich secondary olivine is commonly observed in olivine grains located close to or within these orthopyroxenite veins (Fig. 6f).This replacement has mostly occurred in the rims and along cracks, leaving the olivine cores unmodified (Fig. 6f).

Mineral composition
Mineral composition has been investigated by EPMA in three samples: a garnet lherzolite (JB-19-33), a spinelpargasite-(phlogopite) harzburgite (JB-19-22) and a phlogopite-chromite harzburgite (JB-19-21).A set of selected representative analyses is provided in Tables 2 and 3.The complete set of electron microprobe analyses is provided in the Supplementary Material.
In all samples, olivine displays high Mg# (molar ratio Mg/ Mg þ Fe 2þ ) between 0.88 and 0.91, high Ni (2700-4200 ppm) and moderate Mn (1000-1300 ppm) content, with Ni/Mn between 2.1 and 3.3 (Fig. 7a), which is indicative of mantle peridotite olivine and precludes any cumulative origin for these samples (Wang et al., 2021).This is consistent with the low concentrations in trace elements: below 50 ppm for Ti and Al and below 200 ppm for Ca, Cr and P. In sample JB-19-21 (phlogopite-chromite bearing harzburgite), the rims of olivine grains within or close to the orthopyroxenite layers have lower Mg# (0.81-0.86) and higher Mn content (Mn = 2000-5800 ppm) than the cores, with low Ni/Mn between 0.32 and 1.74.This suggests this olivine population formed by reaction between mantle olivine and melts.Composition variations are observed between samples.In particular, there is a consistent increase in Ni/Mn correlated with increasing Mg# (Fig. 7a), from the least magnesian olivine in clinopyroxenegarnet lherzolites (Ni/Mn = 2.1-2.3) to the more magnesian olivine in clinopyroxene-free phlogopite-chromite harzburgite (Ni/Mn = 3.0-3.3).
Diopside and garnet were analyzed only in the garnet lherzolite.Diopside has high Mg# (0.92-0.94 p.f.u) and variable Al (0.14-0.17 p.f.u Al).Diopside in the coronas (Cpx-II) contains slightly more Al and slightly less Cr than primary diopside Cpx-I, with again no difference in Mg# (Fig. 7d).
Garnet is a slightly chromian pyrope (Cr# = 0.04-0.05).Composition profile (Fig. 7e) reveals a slight Mg-decrease and Feincrease from core to rim (from Prp 73 Alm 15 Grs 11 to Prp 65 Alm 23 Grs 11 )   probably related to diffusion, while no zoning is observed neither in Ca nor in Cr.
6 Whole-rock geochemistry and Nd-Sr isotopes

Major elements
All the samples have been serpentinized to various degrees during retrogression, but this process is generally considered to induce no or very minor changes in major element composition, and preserves in particular Al 2 O 3 /SiO 2 and MgO/SiO 2 ratios (Deschamps et al., 2013).Formation of talc during serpentinization may nevertheless be associated with strong chemical modification, notably SiO 2 enrichment (Paulick et al., 2006).Apart from highly retrogressed or altered domains which have been removed before crushing, no Table 3. Representative EPMA analyses of spinel, chromite and olivine, including trace elements measured with a high intensity electron beam (Batanova et al., 2015).na: not analyzed.The complete set of analyzes is provided in the Supplementary Material.Tableau 3. Sélection représentative d'analyses à la microsonde électronique du spinelle, de la chromite et de l'olivine, incluant les éléments traces mesurés avec une forte intesité du faisceau d'électrons (Batanova et al., 2015).na : non analysé.Le jeu de données complet est disponible en matériel supplémentaire.talc has been observed in the samples.Therefore, the measured compositions re-normalized to an anhydrous basis (Tab. 1) are considered as representative of the whole-rock compositions before serpentinization.Normalizing to dry composition may induce a bias for the phlogopite/pargasite-bearing samples, which were not completely anhydrous before serpentinization.This bias is however assumed to be small considering the very high loss on ignition (8-10 wt%), which is dominated by serpentine and other volatile-bearing phases formed during retrogression rather than pre-existing metasomatic minerals.
Variation in major element composition reflects the mineralogical composition of the peridotites.In particular, the trend in the Al 2 O 3 /SiO 2 -MgO/SiO 2 diagram is consistent with composition of olivine and orthopyroxene, which are more magnesian in the most refractory harzburgites than in the most fertile garnet lherzolite (Figs.7a and 7b).Furthermore, higher CaO and Al 2 O 3 in the lherzolites is consistent with the large abundance ofretrogressedgarnet and clinopyroxene, and higher K 2 O and P 2 O 5 in sample JB-19-21 is consistent with the presence of phlogopite and apatite.

Trace elements
Trace element analyses recast on an anhydrous basis are given in Table 4. Large variability is observed in trace element composition, and three main groups are identified based on REE content (Fig. 9a).The first group includes most of the garnet lherzolites, which present LREE-depleted patterns (La N /Yb N = 0.46-0.96),with HREE content close to or slightly below that of the primitive mantle (Yb N = 0.5-1.1).Samples from this group display no or slightly positive Eu anomaly (Eu/ Eu* = 1.0-1.2,where Eu/Eu* = Eu N / (Sm N .Gd N ) 1/2 ), and are not particularly enriched in Th (Th N /Yb N = 0.48-2.12), in contrast to the other groups.The second group includes the spinel and the phlogopite harzburgites, plus three samples of  (2006).All the samples align slightly below the terrestrial mantle melting array (Jagoutz et al., 1979), possibly due to SiO 2 addition during metasomatism (Paulick et al., 2006) or MgO loss during serpentinization.Larger symbols correspond to samples analyzed by electron microprobe.and Ta, independently of their mineral composition or REE and Th content (Fig. 9b).Th N /Nb N ratio, which quantifies this anomaly, ranges between 1.71 and 54.86.

Sr-Nd isotopes
Nine samples have been selected for bulk-rock Sr-Nd isotopic analyses (Tab.5).However, due to very low Nd content, no Nd isotopic data could be acquired for samples from the LREE-depleted group, and the only available Nd data are from LREE-enriched samples.Initial eNd i have been recalculated at 330 Ma, which corresponds to the climax of the Variscan collision in the ECM.Variations of ± 20 Ma on the initial age have little effect on recalculated eNd i (± 0.07-0.23)compared to the observed variation range.eNd 330 values cover a large range from slightly positive to strongly negative values (À8.12; þ0.59).Slight contamination by granitic melt in the migmatite cannot be excluded, which would shift eNd 330 toward lower values.However, only the freshest part of the enclaves, which did not present macroscopic evidences of melt-rock interaction in the migmatite, has been selected for whole-rock isotopic analyses.Therefore, we consider that the measured whole-rock isotopic compositions are dominated by Fig. 9. Trace element composition of collected samples, normalized to the composition of primitive mantle after McDonough and Sun (1995).Three groups are identified based on their trends in rare earth elements (see description in text).See Figure 8 for the legend.Fig. 9. Composition en éléments traces des échantillons collectés, normalisée à la composition du manteau primitif d'après McDonough et Sun (1995).Trois groupes sont identifiés à partir des spectres en terres rares (voir description dans le texte).Légende identique à la figure 8.   the peridotite component, and are thus representative of the Variscan mantle composition.
In contrast, Sr isotope data are more difficult to interpret.Measured 87 Sr/ 86 Sr ratios range from 0.7057 to 0.7677, with highest values measured in the harzburgites (0.7218-0.7677) and in the serpentinites (0.7173-0.7178).Recalculation of initial isotopic composition at 330 Ma may be flawed by late alteration, in particular in the serpentinites, which present high 87 Sr/ 86 Sr ratios.Furthermore, high Rb/Sr ratios (17 to 35, Tab. 5) in the phlogopite-bearing samples make recalculation of initial 87 Sr/ 86 Sr very sensitive to even small analytical errors on Rb or Sr contents or age variations.Thus, the Sr isotopic compositions cannot be easily interpreted in terms of mantle processes, and we will not discuss these results further.

Thermobarometry
The P-T evolution of the garnet lherzolites has been constrained using inverse thermobarometers based on equilibrium between garnet, clinopyroxene and orthopyroxene.Among the collected samples, only JB-19-33 preserved a complete assemblage with these three phases.P-T conditions have therefore been estimated only on this sample.Two metamorphic stages can be constrained: (i) a first stage in the garnet stability field, which we constrained using the composition of garnet, Opx-I and Cpx-I; (ii) a lower P stage in the spinel stability field, which we constrained using the composition of Opx-II-Cpx-II pairs in the coarse-grained coronas around the kelyphites.We only used composition data acquired in the core of largest grains (ca.200 mm), which are the least affected by post-growth diffusion.As pointed out by Nimis and Grütter (2010), many formulations of thermobarometric equations commonly used for mantle rocks are not internally consistent and may yield biased results, even in wellequilibrated samples.We followed their recommendations and used the Al-in-Opx barometer of Nickel and Green (1985) (NG85) in combination with the two-pyroxene thermometer of Taylor (1998) (TA98) and the Ca-in-Opx thermometer of Brey and Köhler (1990) (BK90) which have proven to yield the most robust results.Furthermore, these thermobarometers do not rely on Fe-Mg partitioning between co-existing minerals, which is particularly sensitive to diffusion resetting (Carlson, 2006;Cherniak and Dimanov, 2010).
Inverse thermobarometry has been completed by thermodynamic modeling with PerpleX 6.9.0 (Connolly, 2009), using the thermodynamic database of Holland and Powell (2011) and the set of solution models from Jennings and Holland (2015) developed for phase relations in peridotite in the composition space NaO-CaO-FeO-MgO-Al 2 O 3 -SiO 2 -Cr 2 O 3 .All iron was assumed to be ferrous (no Fe 3þ ), and the bulk composition of sample JB-19-33 recalculated on an anhydrous basis was used in input.All P-T results are summarized in Table 6.The phase diagram computed with PerpleX is shown in Figure 10a.The primary assemblage composed of olivine, garnet, clinopyroxene and orthoproxene is stable over a large P-T range above 1.3-2.0GPa, while spinel appears at lower P between 0.8 and 1.3 GPa.Equilibration conditions of garnet, Cpx-I and Opx-I has been assessed by computing the composition quality factor Q cmp of Duesterhoeft and Lanari (2020) over the whole P-T grid (Fig. 10b).Q cmp gives the match between modeled and measured mineral composition in function of P and T, using the same mineral compositions as those used for inverse thermobarometry.The closer Q cmp is to 1 (yellow shades in Fig. 10b), the better the match.The best composition match (Q cmp > 0.95) is obtained for P between 2.0 and 4.0 GPa and T between 940 and 980 °C (Fig. 10b).Inverse thermobarometry using the NG85 and TA98 calibrations yields a comparable estimate for the primary assemblage at P = 3.0 GPa and T = 973 °C.Internal calibration uncertainty of these thermobarometers can realistically be assumed to be around ± 0.5 GPa for P and ± 50 °C for T (Nimis and Grütter, 2010) However, assuming the same pressure (3.0 GPa), the Ca-in-Opx thermometer (BK90) yields higher T at 1136 °C.In contrast, Opx-II-Cpx-II pairs in the coronas yield consistent T of 827 ± 50 °C for the TA98 two-pyroxene thermometer and 803 ± 50 °C for the BK90 Ca-in-Opx thermometer, assuming P = 1.0 GPa.
By contrast with the lherzolites, spinel/phlogopite-chromite harzburgites lack of garnet and clinopyroxene, which prevents the use of classical thermobarometers.Moreover, these samples display clear evidence of modal metasomatism, which makes the identification of equilibrium assemblages for thermobarometry more challenging.Therefore, no P-T estimates have been obtained on these samples.Lack of garnet nevertheless indicates equilibration below the spinel/ garnet transition.However, according to Ziberna et al. (2013), this transition may be significantly shifted up to > 5 GPa in Table 6.Summary of P-T estimates in sample JB-19-33.For inverse thermobarometry, reported P and T are average of several measurements.Reported uncertainties of ± 50 °C and ± 0.5 GPa are conservative estimates taking into account composition variation and internal calibration uncertainty (Nimis and Grütter, 2010).Tableau 6. Résumé des estimations P-T sur l'échantillon JB-19-33.Pour la thermobarométrie inverse, les valeurs de P et T reportées, sont des moyennes obtenues à partir de plusieurs mesures dans chaque domaine.Les incertitudes rapportées de ± 50 °C et ± 0.5 GPa sont des estimations conservatrices prenant en compte les variations de composition et les incertitudes internes des différents thermobaromètres (Nimis et Grütter, 2010).

Inverse thermobarometry PerpleX
Assemblage P (GPa) T (°C) TA98 T (°C) BK90 T (°C) P (GPa) 1.0 (assumed) 827 ± 50 803 ± 50 nd 0.8-1.3 Cr-rich depleted harzburgites.Therefore, lack of garnet in the metasomatized harzburgites does not provide any information about initial P, which may well be in the same range as that of the garnet lherzolites.

Discussion
8.1 Chemical overprint of enclaves in the crust The relatively small (< 1 m) peridotite enclaves hosted in migmatites have experienced significant chemical overprint, first at HT conditions in the partially molten crust, and later by low-T fluid addition during retrogression.Precise characterization of this late overprint is needed before discussing more extensively about the mantle composition and mantle processes recorded by these enclaves.Chemical interaction with the migmatite host, either at supra-solidus or sub-solidus conditions, is evidenced by the development of reaction coronas at the contact of the enclaves.However, many enclaves display sharp contact with the migmatites, or are surrounded only by a thin rim of biotite and/or hornblende (Figs. 3a,3b and 3g), which suggests that chemical interactions with the host have been restricted to a thin outer rim, with little effect in the core of enclaves.Migmatite-peridotite interactions are however particularly visible around some harzburgite enclaves, especially on sample JB-19-21, which is surrounded by spectacular radial coronas composed of biotite þ tremolite (Figs.3c and 3d).Furthermore, Opx-rich veins observed in this sample (Fig. 4a) are interpreted as a product of the reaction Ol þ Liq 1 !Opx þ Liq 2, in which primary olivine is dissolved by a relatively SiO 2 -rich melt (Liq 1), producing orthopyroxene and a more mafic melt (Liq 2).These veins could therefore indicate reaction of the peridotite with granitic liquid in the migmatite.However, such reactions only affected centimeterwide domains, as shown by the distribution of olivine grains with Fe-rich rims, which are restrained to the immediate vicinity of Opx-rich veins.Since both the coronas and the metasomatic veins have been carefully removed by cutting before doing geochemical analyses, they should only marginally affect the results of the geochemical analyses.
Serpentinization and other low T retrogression reactions occurred to various degrees in all the collected samples, and resulted in significant addition of volatile components (Loss on Ignition (LOI) up to 13 wt%).However, it does not seem that these reactions have significantly affected the composition of minerals.In sample JB-19-21 (phlogopite harzburgite), Fe-Mg-Mn zoning previously developed in olivine was not modified by the development of the serpentine mesh (Fig. 6f), which suggests little or no effect of serpentinization on olivine composition.Fluid-addition does not seem to be associated either with strong modification of the whole-rock trace element and isotopic composition.Positive anomalies in some large ion elements mobile in crustal fluids (U, Pb, Cs) are observed in all samples, independently of the volatile content in these samples, as shown with U/Th plotted against LOI in Figure 11.Two serpentinite outliers with high LOI however display high U/Th relative to other samples, which might be the result of U addition by high U/Th crustal fluids.Negative HFSE anomalies (Th N /Nb N > 1), LREE enrichment relative to HREE (La N /Yb N ) and Nd content are also not correlated with the LOI (Fig. 11), which suggest that the main geochemical features of the peridotites and serpentinites were acquired before serpentinization.Moreover, the strongly negative anomalies in HFSE correspond to Nb and Ta contents well below the primitive mantle composition for most of the samples (Fig. 9b).Thus, even accounting for possible addition of La, Th and U by crustal fluids, these anomalies likely existed before the chemical overprint in the crust.

P-T evolution of the garnet lherzolites
The primary Grt þ Cpx-I þ Opx-I assemblage in sample JB-19-33 correspond to an initial equilibration stage of the lherzolites in the garnet field, estimated at 973 ± 50 °C and 3.0 ± 0.5 GPa using the NG85 þ TA98 thermobarometers, which is consistent with estimation obtained by forward thermodynamic modeling (2-4 GPa, 940-980 °C).However, there is quite a large discrepancy between the TA98 twopyroxene and the BK90Ca-in-Opx thermometer, which yields T ca. 150 °C difference.This discrepancy cannot be explained solely by internal uncertainties of the respective calibrations, but more probably results also from composition resetting by diffusion after mineral growth.This may explain for instance the relatively large spread in Mg# and Ca observed for clinopyroxene (Fig. 7d, Table S1).In contrast, Al content in Opx presents little dispersion (Fig. 7c), which suggests little or no diffusion has occurred.We therefore consider the Al-in-Opx barometry is quite robust.Taking the TA98 and BK90 as respectively lower and upper estimates for T, and considering the dependence in T of the NG85 barometer, we estimate that the initial equilibration stage occurred around 970-1140 °C and 3.0-4.0GPa.
Formation of kelyphites is indicative of a second metamorphic recrystallization stage at lower pressure.Kelyphitization results from reaction between garnet and olivine in the spinel stability field, following the reaction (Godard and Martin, 2000;Obata, 2011), which occurred between 1.3 and 0.8 GPa according to our thermodynamic calculations (Fig. 10).The coarse-grained Opx-II þ Cpx-II coronas developed concomitantly with the kelyphite (Obata, 2011), and constrain T between 800 and 850 °C during decompression using the TA98 and BK90 thermometers.Finally, the formation of plagioclase-bearing symplectites around garnet relics indicates further decompression in the plagioclase stability field below ca.0.8 GPa.Transformation of the kelyphites into coarser symplectites and eventually into a polygonal assemblage of orthopyroxene þ clinopyroxene ± spinel ± amphibole (Fig. 4b) is a process indicative of static recrystallization at relatively high T conditions > 800 °C (Obata, 2011).Resulting textures are very similar to annealing textures observed in garnet-clinopyroxene HP mafic granulites of the Oisans-Pelvoux Massif (Jacob et al., 2022), which have experienced an overprint at similar P-T conditions (0.6-0.9 GPa; 800-870 °C) during the late Variscan collisional stages .These textures therefore suggest that the lherzolite enclaves have remained in the lower crust at T > 800 °C, following the first stage of decompression that led to garnet breakdown.Fig. 11.Whole-rock (U/Th) N , (Th/Nb) N , (La/Yb) N and Nd N normalized to the primitive mantle composition in function of the loss on ignition (LOI).LOI is considered as a proxy for fluid addition during retrogression in the crust (mostly serpentinization).Lack of correlation with the LOI ((U/Th) N , (Th/Nb) N , (La/Yb) N ) or slightly negative correlation (Nd) suggests that late fluid-addition during retrogression had little effect on whole-rock trace element compositions.Two serpentinite outliers with high (U/Th)N (contoured spots) might have nevertheless experienced significant contamination by fluid.See caption of Figure 8 for the legend.Primitive manle composition after McDonough and Sun (1995).Fig. 11.Sélections de paramètres géochimiques sur roche totale (U/Th) N , (Th/Nb) N , (La/Yb) N et Nd N (composition normalisée au manteau primitif) en fonction de la perte au feu (LOI).Celle-ci est utilisée comme un indicateur de l'intensité des réactions de rétrogression tardives (serpentinisation principalement), associées à un fort apport de fluide.L'absence de corrélation avec la LOI ((U/Th) N , (Th/Nb) N , (La/Yb) N ) ou la présence de corrélation légèrement négative (Nd) suggère que l'apport tardif de fluides pendant la rétrogression a eu peu d'effets sur la composition en éléments traces de la roche totale.Légende identique à la figure 8. Composition du manteau primitif d'après McDonouh and Sun (1995).

A refractory mantle refertilized by melts and fluids
Geochemical and mineralogical investigations of the peridotites and serpentinites reveal significant decoupling between major, trace elements and mineral compositions, which we interpret as the result of metasomatic overprint of variably melt-depleted refractory mantle.Initial melt depletion is inferred from whole-rock mineralogy, Al 2 O 3 /SiO 2 and MgO/SiO 2 compositions, which present large variations from very fertile (high Al 2 O 3 /SiO 2 , cpx-rich) garnet lherzolites to refractory (low Al 2 O 3 /SiO 2 , cpx-poor to cpx-free) harzburgites and serpentinites (Fig. 8).These variations are furthermore consistent with the composition of olivine and orthopyroxene, which display increasing Mg# and Ni and decreasing Mn from the more fertile to the more refractory samples.Finally, even in the most fertile lherzolites, LREE depletion suggests they originally derive from a refractory mantle.
The variable degrees of melt-depletion are however completely decorrelated from the whole-rock trace element compositions.Indeed, largest enrichment in incompatible elements, in particular LREE, Rb, Ba and Th are observed in the refractory harzburgites, while the more fertile lherzolites are depleted in LREE relative to HREE (Fig. 7).Selective enrichment in some LILE (Cs, Rb, U, Pb) relative to HFSE (Nb-Ta) is also observed independently of the mineral and whole-rock composition of the samples, and thus marks a cryptic metasomatic overprint.Some samples also record modal metasomatism, as evidenced by the presence of volatilebearing phases like phlogopite and pargasite in the refractory harzburgites.In the lherzolites, hydrous minerals (pargasite) are also found, but they are mostly located in former garnet sites (kelyphites or polygonal mosaic assemblages), and may be related to retrogression in the crust rather than mantle metasomatic processes.However, the tiny graphite-bearing inclusions found in garnet relics (Fig. 5) suggest the presence of reduced C-O-H fluids during garnet crystallization.Therefore, it is possible that crystallization of garnetclinopyroxene-rich assemblages in the LREE-depleted lherzolites was catalyzed by the presence of these reduced fluids.
Similar decoupling between major and trace elements content in orogenic mantle rocks has been widely documented throughout the Variscan Belt (Scambelluri et al., 2006;Medaris et al., 2015;Kube s et al., 2022).It has commonly been interpreted as the result of an early phase of partial melting, followed by one or multiple phases of metasomatic overprint that led to selective enrichment of the bulk rock in some melt or fluid-mobile trace elements (Medaris et al., 2015;Kube s et al., 2022).The initial refractory character of the mantle rocks may either indicate derivation from a subducted oceanic lithosphere (Kube s et al., 2022), or point toward old (possibly Proterozoic) partial melting events in the sub-continental mantle (Medaris et al., 2015).Decoupling between major and trace element composition may result from limited metasomatic influx, which would only produce selective enrichment in specific elements mobile in the metasomatic phase, leaving the initial major element characteristics of the bulk rock largely unmodified.Such a case has been documented by Borghini et al. (2018) in mantle eclogites from the Granulitgebirge (NW Bohemian Massif).In these rocks, percolation by felsic melts rich in Cs, Li, B, Pb, Rb, Th, and U is evidenced by the presence of tiny crystallized melt inclusions in garnet, while the bulk rock only recorded very limited enrichment in these elements.
Two different metasomatic trends (subsequently referred to as type-I and type-II) are recorded in the Pelvoux mantle enclaves, which variably affected the different samples: all of them present variable enrichment in LILE (Cs, Rb, U, Pb) relative to Nb and Ta (type-I), while only the harzburgites and some lherzolites are enriched in incompatible LREE and Th (type-II).This dichotomy appears more clearly when plotting Th/Nb (amplitude of Nb anomaly) against Th/Yb (enrichment in incompatible elements relative to HREE).The LREEdepleted lherzolite samples are aligned along a very steep trend (Fig. 12), which indicates increasing Nb anomalies at nearly constant Th/Yb (type-I).In contrast, the LREE-enriched samples are generally more scattered and define a much shallower trend (type-II), which indicates significant enrichment in incompatible elements (Th and LREE) relative to more incompatible ones (HREE), without additional fractionation of LILE relative to Nb and Ta.We interpret these two trends as the result of interactions with different types of fluids or melts, which present different affinities with LREE relative to HREE and LILE relative to HFSE.
The signature of type-I metasomatism is very similar to subduction-related metasomatism, which as been extensively discussed in the literature (e.g.Baier et al., 2008;Hermann and Rubatto, 2009;Zheng, 2019).Strong enrichment in LILE relative to Nb and Ta in the metasomatic phase is probably related to melting and/or dehydration at UHP conditions (> 2.5 GPa) of subducted metasediments, in which the stability at HP of phengite, rutile, allanite and monazite exerts a strong control on LILE, Th, LREE and HFSE content of the fluid or melt produced (Hermann and Rubatto, 2009).This type of signature is very common in Variscan mantle (U)HP rocks (Scambelluri et al., 2006;Medaris et al., 2015;Borghini et al., 2018Borghini et al., , 2020;;Kube s et al., 2022), which suggests that metasomatism mainly resulted from percolation by silicic melts released during Variscan subduction.This is furthermore confirmed by Lu-Hf and Sm-Nd garnet ages of some metasomatized garnet peridotites which point toward Devonian-Carboniferous age of metasomatism and UHP metamorphism (Tumiati et al., 2003;Ackerman et al., 2020;Kube s et al., 2022).
Type-II metasomatism presents distinctive characteristics, which can be investigated with more details using mineral compositions and petrographic observations in the phlogopitechromite harzburgite (sample JB-19-21), one of the most LREE-enriched samples (Fig. 9).Petrographic relations in this sample show that metasomatism is associated with coeval crystallization of phlogopite, chromite and apatite.Composition of low-Mg# chromite, which is poor in compatible Ni, and comparatively richer in more incompatible Ti and Mn (Figs. 7g and 7h, Tab.3), suggests it crystallized from a melt phase.We therefore suggest that type-II metasomatism was driven by percolation of the harzburgite by a hydrous silicate melt rich in K, P, Cr and LREE.This signature, which is characterized by simultaneous enrichment in compatible (Cr) and incompatible elements, is typical of ultrapotassic magmas like lamproites or lamprophyres (Foley et al., 1987).It appears that similar ultrapotassic magmas are quite widespread in the Variscan basement of ECM (Debon et al., 1998) and other Variscan Massifs of Western and Central Europe.With their more differentiated counterparts (Mg-K-rich monzodiorite to qtzmonzonite and mela-syenites), they collectively form a magmatic series referred to as durbachites (Von Raumer et al., 2014;Moyen et al., 2017;Soder and Romer, 2018).These durbachites presumably derived from melting of a metasomatized mantle source, enriched by melts derived from a subducted mature continental crust (Janou sek and Holub, 2007;Janou sek et al., 2019Janou sek et al., , 2020)).We therefore speculate that type-II metasomatism is related to percolation in the mantle of the primary magmas from which the durbachites are derived.This hypothesis is supported by experimental work of Foley and Pertermann (2021), who produced a phlogopite harzburgite composition similar to our sample by dynamic percolation of lamproite melt into a garnet peridotite at ca. 3.0 GPa.Finally, the very low eNd i measured in the phlogopite harzburgite sample is also similar to that of durbachites (Janou sek et al., 2019) and confirms involvement of mature continental material in the source of the metasomatic melts.Variable Nd isotopic signatures (À8.12 < eNd i < þ0.59) in the LREE-enriched peridotites suggest that the extent of type-II metasomatism was spatially very heterogeneous, and probably focused along localized melt migration pathways.
Decompression of the garnet lherzolites from > 3 GPa to 0.8-1.3GPa requires significant vertical uplift of at least 50-70 km, assuming lithostatic pressure and a density of 3300 kg.m À3 for the lithospheric mantle.However, the exact mechanism for vertical transport of the mantle enclaves remains unclear.Buoyancy-driven exhumation in a subduction channel, or incorporation as tectonic slices in the downgoing continental slab is commonly suggested to explain the presence of mantle enclaves in crustal units (Guillot et al., 2009;Brueckner et al., 2010;Malatesta et al., 2012).However, the Pelvoux Massif lies hundreds of kilometers south of the main Variscan (Saxo-Thuringian/Rheic) subduction zones (Fig. 1), and there is no clear evidence of a former subduction channel preserved in the ECM (Jacob et al., 2021(Jacob et al., , 2022)).Either a subduction was active in the ECM during the early orogenic stages, and was then dismembered by subsequent tectonics and magmatism, forming a "cryptic suture zone" (Schulmann, Lexa et al., 2014;Jouffray et al., 2020), or such a subduction never existed, and alternative mechanisms have to be involved.In the Moldanubian Zone of the Bohemian Massif, "relamination" mechanisms have been suggested to explain the presence of (U)HP units far from the plate interface.These (U)HP units consist of felsic garnet-kyanitebearing granulites with relics of coesite and/or diamond (Kotková et al., 2011;Perraki and Faryad, 2014), which contain numerous bodies of garnet peridotites and other UHP mantle rocks (Medaris et al., 2005).It is suggested that these units represent buoyant mélanges derived from the slab/mantle wedge interface, which were exhumed by diapiric ascent far behind the subduction front, and eventually incorporated into the overlying lower crust.This relaminated lower crust would have then been locally extruded by up-doming (Janou sek and Holub, 2007; Lexa et al., 2011;Schulmann, Catalán et al., 2014).This mechanism has been corroborated by recent thermomechanical modeling (Maierová et al., 2018(Maierová et al., , 2021)), and could be a viable explanation to the presence of mantle enclaves in the lower crust of ECM, far from the subduction trench.Furthermore, it is rather consistent with (i) the extent of metasomatism observed in the enclaves, and (ii) the sampling of different mantle domains at different depths.However, as opposed to the Bohemian Moldanubian Zone, no evidences of UHP conditions in the felsic crust have been found so far.It is therefore still not clear whether the felsic migmatitic host has experienced such extreme conditions.Recent tomographic reconstruction of the crust and upper mantle of the Western and Central Alps indicates the presence of a low-velocity anomaly in the lower crust below the ECM (Nouibat et al., 2022), which might sign the presence of relaminated felsic material.However, extrusion of this lower crust may have been less efficient than in the Bohemian Massif, which would explain the scarcity of mantle enclaves and the lack of large UHP felsic granulite units in the ECM.
Extensive metasomatic interactions with hydrous melts and/or fluids recorded in the ultramafic enclaves are very likely related to slab/mantle interactions during the Variscan orogeny, although we lack of geochronological data to constrain precisely the age.Indeed, older mantle-derived magmatism in the ECM does not present any evidence of mantle enrichment.In particular, Cambrian-Ordovician tholeiitic magmatism, which is widespread in the ECM, consists only of N/E-MORBs with positive eNd i between þ6 and þ8 indicative of a depleted mantle source (Paquette et al., 1989).Thus, the mantle below the ECM prior to the Variscan orogeny was probably mostly composed of depleted, refractory harzburgitic material.The initial melt-depletion event inferred from whole-rock and mineral compositions of the peridotites may be related to this widespread magmatic event, although older depletion events are not excluded.The isotopic and trace element enrichment observed in the peridotite enclaves was necessarily acquired later, more probably during Variscan subduction and collision.This is in good agreement with the metamorphic and metasomatic evolution of other Variscan orogenic peridotites.Dating of garnet in lherzolites from the Bohemian Massif (Kube s et al., 2022) and from the Ulten Zone (Tumiati et al., 2003) yield indeed Carboniferous ages (ca.340-330 Ma), which points toward a widespread metasomatic event in the mantle occurring during the Variscan collisional stages.This age is very similar to the age of emplacement of the durbachite magmas (Debon et al., 1998), which clearly suggests a connection between this peculiar magmatism and metasomatism observed in mantle rocks.

Conclusion
Enclaves of peridotites and serpentinites discovered in exhumed lower and mid-crustal domains in the Oisans-Pelvoux and Belledonne Massif provide insight into the composition and evolution of the Variscan mantle in the ECM.Petrological and geochemical diversity among the enclaves mark the important heterogeneity of the mantle domain from which they derive.Thermobarometric estimations indicate provenance of these enclaves in the upper mantle wedge or the overlying lithospheric mantle.Refractory compositions, characterized in particular by high MgO/SiO 2 and low Al 2 O 3 /SiO 2 in the whole-rock and high Mg# and Ni-content in olivine, mark an early melt-depletion event, presumably related to the widespread Cambrian-Ordovician magmatism recorded all across the Variscan Belt.Depleted mantle domains were variably metasomatized by a LILE-rich phase, probably derived from a Variscan subduction zone, which produced selective enrichment in Cs, Rb, U, Pb relative to less mobile Nb and Ta.Some samples display in addition modal metasomatism (crystallization of phlogopite and/or pargasite with accessory chromite and apatite), associated with a strong enrichment in LREE relative to HREE.This metasomatism is attributed to percolation by a hydrous LILE-K-P-Cr-rich melt, which may be genetically related with the ultrapotassic magmas of the durbachite series.These geochemical characteristics are in line with wholerock Nd isotopic compositions, which indicate enrichment of the mantle by a continental crust component.Further work is however needed to constrain the precise timing of metasomatic events, understand the exhumation mechanisms of the mantle enclaves and establish more precise correlations with the Variscan magmatic and tectonometamorphic evolution in the ECM.Table S1.Complete set of in situ electron microprobe analyses of minerals.Table S2.Complete set of whole-rock geochemical data, including raw data and data recalculated to an anhydrous basis.

Supplementary Material
The Supplementary Material is available at http://www.bsgf.fr/10.1051/bsgf/2022021/olm.Acknowledgements.This work was supported by the BRGM through the Référentiel Géologique de la France program (RGF).Jean-Baptiste Jacob was funded by a doctoral grant from the ENS de Lyon.We greatly thank Laurent Jolivet and Olivier Vanderhaeghe for editorial handling, and two anonymous reviewers for their useful suggestions and comments, in particular regarding the comparison with Moldanubian mantle rocks.We would like to warmly thank Arnaud Pêcher, who worked about fifty years ago during his PhD and guided Jean-Baptiste Jacob to some key outcrops.We also would like to thank Marianna Jagercikova, who assisted Jean-Baptiste Jacob in the field during the summer 2019, and actually found some of the best-preserved samples.We wish to thank all the technical staff at ISTerre in Grenoble for the preparation of samples and technical help during the analytical sessions.Most of the field investigations were carried out in the central zone of the Ecrins National park, in which collecting rock samples is usually strictly forbidden.We therefore wish to thank the administration staff of the park, who gave us authorization to collect the samples.

Fig. 7 .
Fig. 7. Mineral compositions in selected representative samples from the three different peridotite groups.Note in particular the negative correlation between Mg# and Ni-Mn content in olivine and orthopyroxene on panel (a and b), and the significant difference in trace element composition between high Mg# and low-Mg# olivine in the phlogopite-harzburgite.Distinction between mantle and magmatic olivine is afterWang et al. (2021).Note also the sharp difference in composition between the Mg-Cr-rich mantle-phlogopite in the core of the harzburgite enclave and biotite formed by melt-rock interaction in the coronas (g).Mineral compositions are expressed in atom per formula unit, and Mg# is expressed as the molar ratio Mg/Mg þ Fe 2þ .Fig.7.Compositions minérales dans des échantillons représentatifs trois différents groupes de péridotites.Noter en particulier la corrélation négative entre le Mg# et la teneur en Ni-Mn dans l'olivine et l'orthopyroxène (a et b), et la différence significative de composition en éléments traces entre l'olivine riche en Mg et l'olivine pauvre en Mg dans la harzburgite à phlogopite (JB-19-21).Distinction entre olivine mantellique et magmatique d'aprèsWang et al. (2021).Notez également la forte différence de composition entre les phlogopites mantelliques riches en Mg-Cr dans le coeur de l'enclave de harzburgite et les biotites formées par interaction avec la migmatite dans la couronne réactionelle (g).Les compositions minérales sont exprimées en atome par formule structurale, et Mg# est exprimé comme le rapport molaire Mg/Mg þ Fe 2þ .

Fig. 12 .
Fig. 12. (Th/Nb) N vs. (Th/Yb) N diagram (a), and Nb N vs. La N diagram (b) showing the two metasomatic trends (type-I and type-II) observed in the samples.Compositions are normalized to the primitive mantle of McDonough and Sun (1995), to show the relative evolution of the different groups from this reference.Type-I metasomatism is characterized by increasing HFSE anomaly with little or no fractionation of incompatible elements (LREE þ Th) relative to HREE.This corresponds to an increase in (Th/Nb) N at constant (Th/Yb) N in Figure 12a, and to an increase in La at constant Nb in Figure 12b.In constrast, type-II metasomatism is characterized by strong enrichment in incompatible elements without fractionating fluid-mobile LREE and fluid-immobile HFSE.This corresponds to an increase in (Th/Yb) N in Figure 12a, and to a correlated increase in La and Nb in Figure 12b.According to Paulick et al. (2006), these two trends are indicative of fluid-rock and melt-rock mantle interaction for type-I and type-II metasomatism, respectively.See caption of Figure 8 for the legend.Fig.12. Diagramme (Th/Nb) N vs. (Th/Yb) N (a) et Nb N vs. La N (b), montrant les deux tendances métasomatiques observées dans les échantillons (type-I et type-II).Les compositions sont normalisées par rapport au manteau primitif, afin de montrer l'évolution relative des différents groupes à partir de cette référence.Le métasomatisme de type-I est caractérisé par une augmentation des anomalies en HFSE avec peu ou pas de fractionnement des éléments incompatibles (LREE þ Th) par rapport aux HREE.Cela se traduit par une augmentation du rapport (Th/Nb) N sans modification majeure du rapport (Th/Yb) N (Fig.12a), et par une augmentation de la teneur en La sans évolution notable de la teneur en Nb (Fig.12b).À l'inverse, le métasomatisme de type II est caractérisé par un fort enrichissement en éléments incompatibles, sans fractionnement notable en fonction de la mobilité dans les fluides aqueux (LREE vs. HFSE).Cela correspond à une augmentation du rapport (Th/Yb) N (Fig.12a), et à une augmentation corrélée de teneurs en La et en Nb (Fig.12b).D'aprèsPaulick et al. (2006), ces deux tendances sont les marqueurs respectifs d'interaction du manteau avec un fluide aqueux (type-I) et avec un liquide magmatique (type II).Légende identique à la figure 8.

Fig. 12 .
Fig. 12. (Th/Nb) N vs. (Th/Yb) N diagram (a), and Nb N vs. La N diagram (b) showing the two metasomatic trends (type-I and type-II) observed in the samples.Compositions are normalized to the primitive mantle of McDonough and Sun (1995), to show the relative evolution of the different groups from this reference.Type-I metasomatism is characterized by increasing HFSE anomaly with little or no fractionation of incompatible elements (LREE þ Th) relative to HREE.This corresponds to an increase in (Th/Nb) N at constant (Th/Yb) N in Figure 12a, and to an increase in La at constant Nb in Figure 12b.In constrast, type-II metasomatism is characterized by strong enrichment in incompatible elements without fractionating fluid-mobile LREE and fluid-immobile HFSE.This corresponds to an increase in (Th/Yb) N in Figure 12a, and to a correlated increase in La and Nb in Figure 12b.According to Paulick et al. (2006), these two trends are indicative of fluid-rock and melt-rock mantle interaction for type-I and type-II metasomatism, respectively.See caption of Figure 8 for the legend.Fig.12. Diagramme (Th/Nb) N vs. (Th/Yb) N (a) et Nb N vs. La N (b), montrant les deux tendances métasomatiques observées dans les échantillons (type-I et type-II).Les compositions sont normalisées par rapport au manteau primitif, afin de montrer l'évolution relative des différents groupes à partir de cette référence.Le métasomatisme de type-I est caractérisé par une augmentation des anomalies en HFSE avec peu ou pas de fractionnement des éléments incompatibles (LREE þ Th) par rapport aux HREE.Cela se traduit par une augmentation du rapport (Th/Nb) N sans modification majeure du rapport (Th/Yb) N (Fig.12a), et par une augmentation de la teneur en La sans évolution notable de la teneur en Nb (Fig.12b).À l'inverse, le métasomatisme de type II est caractérisé par un fort enrichissement en éléments incompatibles, sans fractionnement notable en fonction de la mobilité dans les fluides aqueux (LREE vs. HFSE).Cela correspond à une augmentation du rapport (Th/Yb) N (Fig.12a), et à une augmentation corrélée de teneurs en La et en Nb (Fig.12b).D'aprèsPaulick et al. (2006), ces deux tendances sont les marqueurs respectifs d'interaction du manteau avec un fluide aqueux (type-I) et avec un liquide magmatique (type II).Légende identique à la figure 8.

Table 5 .
Whole-rock Sr-Nd isotopic compositions.Initial eNd have been recalculated at 330 Ma, which corresponds to the climax of the Variscan collision in the ECM.Sensitivity of eNdi to variation in initial age is shown by recalculating isotopic compositions at 350 Ma and 300 Ma (values in bracket).Recalculation of initial 87 Sr/ 86 Sr is hampered by high uncertainty related to high Rb/Sr ratios in certain samples, and possible contamination by fluids or melts in the crust.Tableau 5. Compositions isotopiques en Sr-Nd sur la roche totale.Les eNd initiaux ont été recalculés à 330 Ma, ce qui correspond au pic de la collision varisque dans les Massifs Cristallins Externes.La sensibilité de eNdi à la variation de l'âge initial est testée en recalculant les compositions isotopiques à 350 Ma et 300 Ma (valeurs entre parenthèses).Le calcul des rapports 87 Sr/ 86 Sr initiaux est entravé par une forte incertitude, liée à des rapports Rb/Sr élevés dans certains échantillons, et à une contamination possible par des fluides ou des liquides silicatés dans la croûte.Sr 87 Sr/ 86 Sr 147 Sm/ 144 Nd 143 Nd/ 144 Nd eNd 330 eNd