The Champtoceaux Complex (Fig. 1) is located between the South Armorican shear zone and the Nort-sur-Erdre Fault, which coincides with a late Carboniferous coal belt (Sillon houiller de la Basse-Loire). It lies structurally on top of the Mauves Unit and below the Mauges Unit (Fig. 1). The Mauves Unit is constituted by a thick, a priori monotonous, series of metagraywackes (albitic micaschists) whose protoliths are supposed to be Proterozoic (e.g. Ballèvre et al., 2009). The Mauges Unit (in the upper allochthonous position) is made up of a Proterozoic basement (micaschists, amphibolites and metavolcanics), on which a Palaeozoic sedimentary succession lies unconformably. The latter begins either in the Cambrian (Cléré-sur-Layon) or in the Lower Ordovician (Châteaupanne). After the regional geological works of the 19th and 20th centuries, Jean Cogné reinterpreted in 1966 the Champtoceaux Complex as being a deep-rooted “nappe” (Cogné, 1966), made essentially of high-grade metamorphic rocks, in which a set of overlapping slices were subsequently recognized (e.g. Ballèvre and Marchand, 1991; Ballèvre et al., 1994; Fig. 1). The composite middle allochthonous units include the Folies-Siffait Unit (amphibolites and metaperidotites, of oceanic affinity) and the Drain Unit (peridotites and metagabbros, recognized as oceanic). The Champtoceaux Unit, made up of migmatitic orthogneisses probably derived from Cadomian protoliths, migmatitic metapelites and rare lenses of eclogite, is structurally situated between the two oceanic units. The eclogite-facies rocks were exhumed and cooled before being involved in the formation of the final Champtoceaux nappe pile associated with the development of inverted metamorphic zoning (Pitra et al., 2010). The lower slices (lower allochthonous) are mainly composed of leucocratic orthogneiss (“leptynites”) derived from Ordovician granitoids (Saint-Mars-du-Désert orthogneiss: Paquette et al., 1984; Cellier orthogneiss: Ballèvre et al., 2002) and metapelites (micaschists) of unknown age, but which are probably also Palaeozoic.

The Cellier Unit is the lowest slice of the Champtoceaux Complex and is mainly made up of an orthogneiss whose transition with an undeformed metagranite can be occasionally observed (Lasnier et al., 1973) and which encloses numerous metre-sized fine-grained mafic eclogite lenses and very few occurrences of metahornfels. The mafic eclogites were first described scientifically as “eurites grenatiques” by François Dubuisson (1830). They were studied by Lacroix (1891) and Brière (1920) at the same time as the eclogites from the Essarts Unit. Lacroix noted that, unlike the latter, the eclogites from Champtoceaux were generally very fine grained and formed small lenses from 1 m to a few metres, stretched in strongly deformed gneisses. Petrological, microstructural and geochronological studies were then undertaken on the eclogites (Velde, 1970; Godard et al., 1981; Godard, 1988; Bosse et al., 2000) and the surrounding metagranitoids and orthogneisses (Lasnier et al., 1973; Paquette et al., 1984; Ballèvre et al., 2002). The eclogites contain relics of doleritic textures, and their composition does not show such marked tholeiitic differentiation as in the several-kilometre-sized eclogite bodies of the Essarts Unit, located about 50 km to the southwest (Bernard-Griffiths and Cornichet, 1985; Godard, 1988). It therefore seems that the protolith of these rocks could have been mafic dykes in a thinned granite-rich continental crust (Ballèvre et al., 1994). The metahornfels, well exposed near Campbon particularly at Le Padé quarry (Marchand et al., 1989), are fine-grained massive rocks with biotite, plagioclase and pseudomorphs after cordierite, in which garnet + phengite + rutile + quartz reaction coronae have been observed, similar to those observed in the Cellier metagranite to which they are genetically related (Godard, 2009). An increasing grade of the eclogite-facies P–T conditions from east (La Varenne; 1.5–2.0 GPa, 550 ∘C) to west (Fay-de-Bretagne; 2.0–2.5 GPa, 650 ∘C) was documented in the Cellier Unit (Ballèvre and Marchand, 1991). This HP metamorphism was dated at ∼ 360 Ma (Bosse et al., 2000; U–Pb zircon and Sm–Nd garnet dating).

This metagranite was first described by Lasnier et al. (1973) as an undeformed core of some wrongly labelled “charnockitic” granite showing, at its rims, a gradual transition to the strongly deformed orthogneiss in which it is embedded. They attributed this metamorphism to a pre-Cadomian granulite-facies process occurring at high temperature while describing the garnet-forming coronitic reaction between biotite and primary plagioclase, as well as the transformation of the latter into a pseudomorph of neo-plagioclase, white mica and clinozoisite. The discovery and studies of the jadeite-bearing metagranite of Monte Mucrone (Italian Alps; e.g. Compagnoni and Maffeo, 1973) and similar rocks metamorphosed under HP conditions, as well as the vicinity of mafic eclogites, led to the reconsideration of this rock as an HP eclogite-facies metagranite (Girardeau, unpublished data). Dating by the Rb–Sr whole-rock method yielded an errorchron of 570 ± 110 Ma, whereas the U–Pb zircon method provided a lower intercept on concordia at 423 ± 10 Ma (Vidal et al., 1980).

At the outcrop scale (Fig. 1b), this undeformed granite appears as 10 m sized boudins, which change on some decimetres to blastomylonitic orthogneiss by the intermediary of increasingly stretched augen gneiss (Lasnier et al., 1973; Cavet et al., 1978). The foliation is concordant with that of the enclosing orthogneiss. The studied samples were taken from boulders of undeformed granite, several metres in size, cropping out at 47∘21′1.5′′ N, 1∘25′41.6′′ W. Numerous loose blocks of eclogite are visible in the vicinity, one of which was chosen for the petrological study (PI2105: 47∘21′6.1′′ N, 1∘25′42.8′′ W).

The studied samples were observed under a scanning electron microscope (SEM) at Paris Sorbonne Université, producing in particular back-scattered-electron (BSE) images and X-ray elemental maps, which were processed using the program XMAPTOOLS 2.2.1 (Lanari et al., 2014).

Electron microprobe (EMP) analyses were carried out at CAMPARIS (SU-IPGP, Paris, France), using CAMECA SX-100 and SX-Five instruments with the data-reducing method of Pouchou and Pichoir (1991). Analytical conditions for spot analysis were 15 kV accelerating voltage and 10 nA specimen current with a beam diameter of 2 µm. Fe2O3, MnTiO3, garnet, Cr2O3, Ba2SiO4, K-feldspar, anorthite, albite and zircon were used as standards. Special conditions of 25 kV accelerating voltage and 10 nA specimen current were used to analyse ZrO2 in rutile. To estimate the bulk composition of the pseudomorphs after plagioclase, representative areas were analysed by integrating contiguous areas (15 µm × 15 µm) scanned by the electron beam. Representative microprobe data are presented in Table 1.

The mineral compositions obtained were used to balance the hypothetical metamorphic reactions deduced from the observed microstructures, using an algorithm and freeware based on the least-square method (e.g. Godard, 2009; Adjerid et al., 2015). This algorithm provides a vector equation, ∑ni (Pi)+(R)=(0), where vectors Pi are the molar compositions of the i phases involved in the reaction, scalars ni are the corresponding stoichiometric coefficients, and R is the residual vector resulting from the least-square regression. One ni is arbitrarily fixed and the others are unknown. Each coordinate of vectors Pi and R corresponds to the content of one independent chemical component. The balance of a reaction is considered satisfactory when (a) the results account for the observed microstructures, with the disappearing and appearing phases as reactants (ni<0) and products (ni>0), respectively, and (b) the residuals (i.e. the coordinates of vector R) are low, a condition fulfilled when the reaction is almost isochemical.

Two samples, a mafic eclogite PI2105 and a metagranite PI2102, were sent for bulk-rock chemical analyses to ALS geochemistry (Loughrea, Ireland). Sample preparation consisted of fusion with LiBO2 and dissolution in HNO3. Major elements were analysed by inductively coupled plasma atomic emission spectroscopy (ICP-AES), and trace element concentrations were determined by inductively coupled plasma mass spectrometry (ICPMS).

The bulk compositions of eclogite PI2105 and metagranite PI2102 are presented in Table 2. The same algorithm used to balance the reactions was also applied to calculate the modal proportions of the minerals in each rock, taking into account the molar compositions of the bulk rock (P1) and of the minerals (Pi>1). The result is considered satisfactory if the residuals are low and the minerals are “produced” from the rock, i.e. if the calculated ni>1 values are all positive, whereas the coefficient n1 for the bulk-rock composition is fixed to -1. The stoichiometric coefficients were converted into volume percentages (Table 2) using the molar volumes given by Abers and Hacker (2016).

The main problem that arises during the thermodynamic modelling of rocks like the metagranite PI2102 is that equilibrium has not been reached at the scale of the rock during metamorphism, since much of the magmatic minerals, still preserved, remained metastable (e.g. K-feldspar, biotite, ilmenite). However, local equilibrium was approached in microdomains, at the scale of the reaction coronae that have grown at the interface of these minerals, so that it is required to model each microdomain instead of the bulk rock. One way to do this is balancing the metamorphic reaction that produced a specific corona by the least-square method. Ideally, if the residuals are low, the overall composition of the products is virtually equal to that of the reactants, representing the atoms that have moved from reactants to products during the reaction and thus can be used as an effective chemical composition to build an isochemical P–T pseudosection (e.g. Godard, 2009; Adjerid et al., 2015). Unfortunately, two complications may arise: (a) one reactant has completely disappeared during the reaction – this is the case of the magmatic plagioclase of the metagranite studied; since its composition is unknown, the reaction cannot be accurately balanced; (b) the residuals for some components are significant, indicating that the system was open to these components during the reaction – this is particularly the case for the reaction at the K-feldspar–biotite interface during which some K2O is released. For these reasons, we opted for the measurements by image analysis of the overall composition of the products. For each reaction, the volumetric proportions of the corona minerals were estimated by image analysis of SEM chemical maps and converted into molar proportions thanks to the molar volumes given by Abers and Hacker (2016). The proportion of the products combined with their average composition obtained at the EMP allowed us to calculate an overall composition of the products, considered as the local effective chemical composition for modelling. A feedback control was then carried out by checking that the composition obtained was also a linear combination of the reactants, either preserved (biotite, K-feldspar, and/or ilmenite) or disappeared (plagioclase of stoichiometry Abx An1-x).

Since such problems do not arise for the eclogite sample PI2105, which is homogeneous and indeed without relict igneous minerals, an effective chemical composition was estimated (SiO2 50.80, TiO2 1.29, Al2O3 9.75, FeO 8.28, Fe2O3 0.22, MnO 0.14, MgO 12.79, CaO 13.31 and Na2O 3.42 in mol %) from the bulk-rock composition. Garnet crystals display well-defined zoning with distinct cores. Considering a spheric geometry, the relative composition of the core and rim of the garnet is estimated thanks to microprobe transects and subtracted from the bulk composition (considering all Mn is in the garnet) to infer peak equilibrium conditions. The small quantities of P2O5 and K2O and the corresponding small amounts of apatite and muscovite were subtracted from the bulk composition. The quantity of Fe2O3 was estimated by expressing the bulk-rock composition as a linear combination of the average compositions of the peak minerals, the Fe3+ amount of which was estimated by stoichiometry (clinozoisite: Fe3+= Fetotal; omphacite: Fe3+ giving 4 cations for 6 O; garnet and amphibole: Fe3+ negligible).

Thermodynamic modelling was performed using the Perple-X software (version 6.8.4; Connolly, 1990, 2005). In the absence of carbonates and negligible presence of organic matter in the studied sample, CO2 was neglected and a fixed water activity of 1 was used for the fluid. The thermodynamic dataset of Holland and Powell (2003) was used with the following set of activity models for solid solutions: garnet, epidote (Holland and Powell, 1998), chlorite (Holland et al., 1998), omphacite (Green et al., 2007), feldspars (Holland and Powell, 2003), white mica (Coggon and Holland, 2002), biotite (Powell and Holland, 1999) and amphibole (Diener et al., 2007; only in sample PI2105).

The bulk-rock composition of PI2102 indicates it is a peraluminous (Al / (Na + K + Ca) >1 in mole ratio) granite (SiO2=70.5 and Na2O + K2O = 7.51 in wt %) following the classification of Shand (1943) and Le Bas et al. (1986). The protolith was a typical granite composed of quartz, biotite (XMg=0.31–0.47; Ti = 0.21–0.29 atoms per formula unit, apfu), potassic feldspar (Ab5-35 An0 Kfs65-95) with albitic perthite lamellae (Fig. 2a) and plagioclase now completely replaced by a microcrystalline albite-rich pseudomorph, in which antiperthitic lamellae are still preserved (Fig. 2b) and show an enrichment in Ba (Ab7 An0 Kfs88 Cls5). K-feldspar is now microcline but shows the Carlsbad twinning inherited from the magmatic orthoclase. The average composition of the initial magmatic plagioclase was estimated (An22 Ab78) thanks to the mass-balancing algorithm considering the whole-rock composition as a linear combination of biotite + quartz + K-feldspar + albite + anorthite + ilmenite + monazite + zircon. This combination is used in the Streckeisen diagram (Streckeisen, 1976) to identify the initial rock of PI2102 as a monzogranite (Q (quartz) 37.1, A (alkali feldspar) 28.0 and P (plagioclase) 34.9, in vol %). Ilmenite, xenotime, monazite and zircon are the main accessory minerals.

The “rapakivi” texture, consisting of a millimetre-thick plagioclase mantling a centimetre-sized ovoid K-feldspar, has been observed once and was also previously described by Lasnier et al. (1973). Several mechanisms can explain this texture (i.e. depressurization or magma mixing); as it has been observed only once and since other K-feldspar crystals are unmantled, it is better explained by magma mixing with the transfer of mantled crystals formed in a hybrid rock to the granite (Vernon, 2016).

Several transformations took place during metamorphism: the replacement of plagioclase and the formation of reaction coronae at the biotite–plagioclase, biotite–K-feldspar and ilmenite–plagioclase interfaces. Quartz is never reacting; however, phengite and garnet replacing biotite may occur at the contact with quartz but always in the continuation of a corona between biotite and feldspar (Fig. 5b).

The bulk-rock composition of PI2105 indicates that its protolith is a basalt or gabbro (SiO2= 48.2 and Na2O + K2O = 3.25 in wt %; Le Bas et al., 1986) with a composition close to N-MORB but with a relative enrichment in Th (Th / Yb = 0.13 and Nb / Yb = 0.64), which identifies it as a C-MORB following the classification of Pearce (2008). This composition indicates a continental contamination which is compatible with the previously suggested origin of this eclogite as mafic dykes in a thinned granite-rich continental crust (Ballèvre et al., 1994). The mafic eclogite PI2105 (Fig. 9) is composed of omphacite (XMg= 0.81–0.92; XNa= 0.42–0.50), clinozoisite (Fe3+= 0.05–0.14 apfu), garnet, amphibole, quartz, paragonite and rutile. Garnet contains inclusions of rutile in both core and rim as well as clinozoisite, omphacite, quartz, and few inclusions of ilmenite and sodic-calcic amphibole in the core (taramite; XMg= 0.32–0.35; XNa= 0.45–0.47). Garnet grains are zoned with a decrease of spessartite content from core (Fig. 4c; Alm46 Prp20 Sps11 Grs23) to rim with an intermediate Fe-rich mantle (Fig. 4c; Alm63 Prp10 Sps1 Grs26) and an Mg-rich rim (Fig. 4c; Alm47 Prp30 Sps0 Grs23). Garnet crystals are in textural equilibrium with omphacite, clinozoisite, quartz and rutile, which likely represent the peak paragenesis.

The retrograde assemblage is characterized by plurimillimetric poikiloblastic amphibole porphyroblasts that grew at the expense of the peak paragenesis and enclosed garnet, omphacite and clinozoisite (Fig. 9b, c). The occurrence of amphibole porphyroblasts in the eclogite of the Cellier Unit has been long described since Lacroix (1891), and they were identified as glaucophane in some eclogites (Brière, 1920; Godard et al., 1981) and as brown amphibole, suspected to be barroisite, in others. In the studied eclogite, it is a barroisite (XMg= 0.86–0.90; XNa= 0.38–0.42) and may present local magnesio-hornblende composition (XMg= 0.78–0.80; XNa= 0.28–0.30) near some clinozoisite or omphacite due to local equilibrium. Balancing the barroisite-forming reaction (see method in Sect. 3.1) yields the following results:

1.220 Grt + 2.564 Omp + 0.949 Qz + 0.010 Rt + 1.493 H2O → 1.000 Amp + 0.501 Czo + 0.242 Pg + 1.834 substitution Fe+1Mg-1 (in molar quantities with almost null residues), which, after transformation into volume amounts, gives a slight increase in volume,

0.53 Grt + 0.60 Omp + 0.08 Qz + 0.00 Rt → 1.00 Amp + 0.25 Czo + 0.13 Pg (volumes; ΔVsolid/Vsolid=+0.119).

Quartz is apparently the limiting reactant, since it no longer exists as inclusions in the barroisite porphyroblasts (Fig. 9c). Clinozoisite, which predates the formation of barroisite, was partly formed during this reaction and incorporated into the amphibole porphyroblasts in the form of numerous inclusions. Paragonite (Fig. 4b; Prl0-4 Ms0-12 Cel0 Pg85-100) also occurs in the retrograde assemblage, as well as in a few occurrences with taramite in inclusions in the garnet core. Finally, it should be noted that plagioclase is not present among the products of the reaction, which suggests that the latter, although clearly retrograde, took place under relatively HP conditions.

The P–T pseudosection for the mafic eclogite PI2105 (Fig. 10a) was calculated in the NCFMASHTO system with excess water owing to the abundance of hydrated phases (large amount of Czo, Amp and Pg). The P–T conditions for the formation of the retrograde barroisite porphyroblasts are constrained around 640 ∘C and 2.1 GPa by the isopleths XNa (0.39–0.42) and XMg (0.86–0.90) of this amphibole (Fig. 10a). The peak P–T conditions are further constrained around 2.0–2.2 GPa and 640–690 ∘C, based on the XMg (0.85–0.89) of omphacite and the Grs fraction (0.23–0.26) of garnet. Under these conditions, the compositions of omphacite (XNa= 0.46–0.47) and garnet (Alm41-42 Prp31-33 Sps0.8) match satisfactorily those obtained with the electron microprobe except for a slightly higher pyrope fraction (>30). These estimated P–T conditions fall within the field Czo + Grt + Omp + Rt + Qz, with predicted modal (vol %) abundances of 43 %–44 % garnet, 2 %–3 % clinozoisite, 50 % omphacite, 1.3 % rutile and 2 %–3 % Qz. The fact that the stability field of the retrograde amphibole is adjacent to the peak P–T field and that the modal abundance of amphibole drastically increases with decreasing temperature suggests an early retrogression of the peak paragenesis (Czo + Grt + Omp + Rt + Qz) with the formation of the large barroisite porphyroblasts at still HP eclogite-facies conditions (∼ 2.0 GPa), outside the plagioclase stability field.

To clarify the evolution of the pseudomorph after plagioclase, a P–T pseudosection was calculated for the theoretical average composition of the magmatic plagioclase (Fig. 10b; An22 Ab78), although the pseudomorph is slightly depleted in alkalis compared to a plagioclase composition, suggesting some alkali mobility during metamorphism (see Sect. 4.1.1). The calculation was performed with an excess of water, given the relatively large amount of clinozoisite and mica in the pseudomorph. The P–T field corresponding to the pseudomorph assemblage (i.e. albite-rich plagioclase + paragonite + clinozoisite + quartz) is found at pressure below the jadeite destabilization curve (green area in Fig. 10b). The peak conditions, estimated from the mafic eclogite, would correspond to the jadeite + paragonite + clinozoisite + quartz field. A lower water content would not change the picture, as it would also produce a peak assemblage with kyanite and/or grossular, associated with jadeite, which are not observed in the pseudomorph.

Estimating the P–T conditions for the formation of the various coronae in the metagranite is arduous because equilibria are only local, with important chemical gradients and a potentially high mobility of some components such as water and alkalis. For example, Schorn (2022) documented large Fe, Mg and Ca gradients in coronae of HP metagranite, resulting in garnet zonations similar to those observed in PI2102. By equilibrating the corona formation reaction at the biotite–K-feldspar interface in a similar rock from the Essarts Unit, Godard (2009) also showed that excess K2O was produced and likely mobilized by fluids. We therefore focused on local systems that were close to isochemical, except for H2O. The coronae at the biotite–plagioclase interfaces obey this criterion. On the other hand, balancing the reactions between biotite and K-feldspar (also featuring an asymmetric albitic rim) and between ilmenite and plagioclase yields high residues of Na2O and K2O, suggesting their mobility, and this reaction was thus not considered.

Regarding the corona between ilmenite and plagioclase, even if the destabilization of ilmenite to rutile is an indicator of HP metamorphism (e.g. Angiboust and Harlov, 2017), the absence of phengite among the products makes it difficult to estimate whether the corona formed at a higher pressure than the neo-plagioclase rim (An18 Ab82), which yields the same P–T stability field as the pseudomorph after plagioclase and may thus be retrograde. The analysis of Zr in the rutile replacing ilmenite provided an average concentration of 68.1 ± 1.1 µg g-1 (± 2 SE). Following the pressure-dependent thermometer of Kohn (2020), the corresponding temperature is T(∘C) = 39.6 × P(GPa) + 473.5, which yields temperature in the range of 533–572 ∘C between 1.5 and 2.5 GPa. This temperature is significantly lower than the peak conditions estimated from the mafic eclogite and potentially indicates either a retrograde or prograde replacement of ilmenite or a local Zr undersaturation despite the existence of large zircon crystals in the metagranite, meaning that the calculated temperature would therefore represent a minimum crystallization temperature.

A P–T pseudosection was modelled for the coronae at the biotite–plagioclase interface (Fig. 10c). The modal amounts of the products, measured by image analysis of Fig. 5c and converted into molar proportions (0.225 Grt + 0.469 Pl + 0.152 Phg + 0.097 Qz + 0.057 Rt), yielded an effective chemical composition of SiO2 57.61, TiO2 1.54, Al2O3 14.56, FeO 11.15, Mn 0.28, MgO 2.50, CaO 2.65, Na2O 4.74 and K2O 1.62 (mol %). This composition seems to have varied little during the reaction since it produces relatively low residues (SiO2 -0.003, TiO2 0.000, Al2O3 0.051, FeO -0.004, MgO -0.006, CaO -0.025, Na2O -0.043 and K2O -0.047 in molar amounts) when compared to the reactants (1 An22Ab70+ 0.421 Bt + 0.159 substitution Fe+1 Mg-1) (see the mass-balance algorithm in Sect. 3.1). The mass balancing also indicates that the corona formation produced water (+0.191 H2O), which is in accordance with Schorn (2022), who suggests that the partial breakdown of biotite produced enough water to hydrate all the pseudomorphs after plagioclase in eclogite-facies metagranites. We therefore considered water in excess. The Si content of the potassic white mica is compatible with the potentially retrograde conditions estimated for the Ab + Phg + Grt + Rt + Qz field, which corresponds both to the pseudomorph after plagioclase and to the actual mineral composition of the coronae. This Si content is also compatible with the peak conditions estimated for the mafic eclogite, which lie in the jadeite stability field (Omp + Phg + Grt + Rt + Qz; Fig. 10c), whereas this mineral was not observed in the metagranite. Modelled garnet composition (Alm70-75 Prp11-14 Sps2 Grs10-18) in the area with Si-rich phengite (Si = 3.20–3.23 apfu) is similar to the composition obtained with the electron microprobe for the dominant almandine-rich garnet. However, the isopleths of garnet were not taken into account as they largely depend on the Ca, Fe and Mg chemical gradients. Lower water content would result in an increase in the stability of kyanite (not observed) and imply a similar trend of the Si content of phengite compatible with both peak and retrograde conditions. Here, we tentatively model the coronae formation by making several assumptions on the local effective chemistry, and the results must be taken with caution. This qualitative result nonetheless corroborates a formation at relatively high pressure (>1.5 GPa) even if we show that the compositions of garnet (mainly driven by diffusion) and phengite (Si content not acting as a barometer) cannot be used for further constraints.

To check if the metagranite reached the same peak conditions as the mafic eclogite, a pseudosection (Fig. 10d) was modelled for the kyanite-bearing xenolith (Sect. 4.1.5), the composition of which is not controlled by local diffusion, contrary to the coronae. The local composition (SiO2 46.75, TiO2 0.002, Al2O3 18.16, FeO 13.13, Mn 1.05, MgO 7.88, CaO 0.17 and K2O 3.99 in molar amounts) was estimated by image analysis in a central zone (Fig. 8c), relatively distant from the inherited garnet of Fig. 8d and from the xenolith border where albite and Ca-rich garnet were likely produced by local chemical contamination from the surrounding pseudomorph after plagioclase. Water was considered in excess because of the large amount of biotite and white mica. The results (Fig. 10d) indicate that an area around 560–610 ∘C and at P>1.7 GPa in the Grt + Chl + Ky + Bt field fits the local Si content of phengite (3.21–3.25 apfu) and the garnet composition (Grs1 Sps6 Alm67 Prp26). Predicted modal abundances (vol %) of ∼ 56 % phengite, ∼ 34 % garnet, 6 %–7 % chlorite and 3 %–4 % kyanite agree with observations. The estimated peak conditions of the mafic eclogite are lying in a field of Mg-rich biotite + phengite + garnet + kyanite, which corresponds to the actual mineral composition of the xenolith with only minor chlorite and phengite partly replaced by biotite. This suggests that the estimated paragenesis of the xenolith corresponds to the late prograde stage of the metagranite.

As discussed in the previous section, estimating the precise P–T conditions of the formation of the coronae is difficult because the composition of their minerals mainly depends on local diffusion gradients (e.g. Ca, Fe, Mg) and on the amount of H2O available. We have also shown that the Si content of white mica is poorly constraining (Fig. 10c). Furthermore, corona formation may occur over a large range of P–T conditions, as evidenced by Bruno and Rubbo (2006) for similar metagranitoid of the Monte Mucrone Massif where the garnet coronae would have formed over a range of 0.2 GPa and 100 ∘C during 2 Ma. Finally, in our case, minerals indicative of clear HP conditions, like jadeite or kyanite, either did not form or were not preserved during retrogression. Nevertheless, the presence of a kyanite-bearing xenolith confirms that the metagranite underwent a pressure higher than 1.7 GPa, which is consistent with the pressure estimated for the nearby mafic eclogite between 1.95 and 2.2 GPa. A slight discrepancy in pressure between the granite and the mafic lenses may be explained by their differences in strength or block aspect ratio (e.g. Moulas et al., 2014; Luisier et al., 2019; Bauville and Yamato, 2021). At such eclogite-facies pressure condition, plagioclase should be unstable with its calcic component transforming into garnet/zoisite + kyanite + quartz ± diopside and its sodic component into jadeite + quartz ± paragonite, depending on the water saturation degree. Neither omphacite/jadeite nor kyanite are observed in the coronae and the pseudomorph after plagioclase of the La Picherais metagranite. The estimated peak P–T conditions are similar to those made in the western part of the Cellier Unit (Fay-de-Bretagne; 2.0–2.5 GPa, 650 ∘C) and are higher than the estimates made in the east (La Varenne; 1.5–2.0 GPa, 550 ∘C), which corroborates an increasing grade of the eclogite-facies event from east to west (Ballèvre and Marchand, 1991; Fig. 11).

The Cellier Unit underwent a polycyclic metamorphism (Fig. 11), with a first high-temperature (HT) event consisting of granite intrusion (423 ± 10 Ma; Vidal et al., 1980) and formation of hornfels (e.g. Le Padé; Figs. 1, 3), followed by cooling and then the HP cycle around 360 Ma (Bosse et al., 2000). The first cooling stage is corroborated by the occurrence of antiperthitic exsolution before the HP event with presence of antiperthitic lamellae preserved in the pseudomorph after plagioclase (Fig. 2b). The HP metamorphism affected an already structured thinned continental crust with mafic dyke swarm being eclogitized in the same event as the surrounding felsic rocks, mainly metagranitoids but also metahornfels (Ballèvre et al., 1994; Godard, 2009). The retrograde P–T path, reported very imprecisely in Fig. 11, is little constrained. However, the fact that the omphacite crystals are well preserved, being only replaced by cryptocrystalline symplectites at their very edge, as well as the growth of glaucophane during the retrogression of some eclogites from the Cellier Unit (Godard et al., 1981) suggests a relatively cold exhumation path (Fig. 11), which also seems corroborated by the relative positions of the P–T conditions at the peak and during the barroisite growth in the studied mafic eclogite PI2105 (peak and retro, respectively, in Fig. 10a). Finally, in order to preserve the mineral assemblage that replaced plagioclase in the metagranite, this retrograde P–T path must have followed its stability field (“Pl psd” in Fig. 11) down to fairly low temperatures.

A first comparison can be made with the Malpica–Tuy Unit in the Iberian Massif in NW Spain, which exposes one of the best sections of the Ibero-Armorican Arc (Martínez Catalán et al., 1996; Catalán et al., 2007; Ribeiro et al., 1990). Malpica–Tuy is structurally and petrologically equivalent to the Cellier Unit and is part of the lower allochthon. As in the Armorican Massif, it consists of micaschists and fine-grained orthogneisses derived from Ordovician granites/rhyolites and contains both well-preserved eclogite and poorly deformed metagranitoid boudins. Protolith (∼ 495 Ma: Abati et al., 2010) and HP metamorphism (365–370 Ma: Rodríguez et al., 2003; Abati et al., 2010) ages are similar, and P–T conditions (>1.6 GPa; 640 ∘C for the metagranitoids; Gil Ibarguchi, 1995) are also equivalent to the estimates we obtained for the La Picherais metagranite. These two metagranites are similar in all respects according to the description of Gil Ibarguchi (1995): (1) garnet coronae formed between biotite and plagioclase with a rim of albite ± phengite on the plagioclase side and with partial replacement of biotite by white mica, which in turn is partly replaced by secondary biotite; (2) garnet coronae can be found between ilmenite and plagioclase with ilmenite being partly or totally replaced by rutile; and (3) pseudomorphs after plagioclase composed of zoisite, albite and phengite are observed. Unlike our observations, Gil Ibarguchi (1995) documented the presence of jadeite in one locality where barely deformed jadeite-bearing rocks consist of 1 to 10 cm thick porphyritic and aplitic veins that crosscut hornfelsic metasediments enclosed in the orthogneiss. In these veins, jadeite occurs in pseudomorphs after plagioclase but mainly near biotite inclusions, suggesting preferential nucleation of jadeite along grain boundaries. The lack of jadeite in inner parts of pseudomorphs after plagioclase and the irregular development of corona textures indicate fluid restrictions and nucleation barriers during HP metamorphism (Gil Ibarguchi, 1995).

Another comparison should be made with the eclogite-facies gneiss of the Essarts Unit, which is located south of Nantes and of the South Armorican Shear Zone, at barely 50 km from La Picherais. The Essarts Unit is mostly made of continental crust composed of orthogneiss and paragneiss, but it also includes kilometre-scale boudins of mafic eclogites and altered garnet peridotites considered as remnants of a subducted oceanic lithosphere (Godard, 2001, 2009). As in the Cellier Unit, the gneisses underwent two distinct high-grade metamorphic cycles (Godard, 2009), with a first HT and low-pressure event that produced partial melting in the cordierite stability field (migmatitic paragneiss; T≈650 ∘C, P≈0.4 GPa) accompanied by intrusions of granitoids (now orthogneiss) and followed by a later HP eclogite-facies overprint (T≈700 ∘C, P>1.8 GPa; Godard, 2009). In the ortho- and paragneisses, the HP metamorphism is marked by many pseudomorphic and coronitic reactions, which is more varied than at La Picherais because other HT minerals such as cordierite are involved. Nonetheless, they show similar coronae between biotite and plagioclase (phengite + rutile / garnet + rutile / garnet + quartz / albite), between ilmenite and plagioclase (rutile / garnet / garnet + quartz / albite), and between biotite and K-feldspar (garnet / garnet + quartz / quartz + phengite). The grossular content of garnet also increases near the pseudomorph after plagioclase. A difference with La Picherais is the presence of some kyanite in the pseudomorphs after plagioclase composed of albite–kyanite–zoisite–phengite with kyanite microcrystals delineating polygonal millimetre-sized cells with a honeycomb-like structure that has been interpreted as inherited from the shapes of the HT plagioclase grains. Godard (2009) suggested that the pseudomorphic recrystallization of millimetre-sized albitic single crystals into polycrystalline 10 µm sized albite crystals is likely the result of a phase transition process such as albite to jadeite, which would imply that jadeite crystallized at high pressure before being totally retrogressed into albite (no relic found). Finally, phengite is partly replaced by a late biotite II during the final retrogression. Further south, other Variscan occurrences of gneiss and metagranitoids associated with mafic eclogites and showing similar HP coronae have also been reported in the French Massif Central (see review in Godard, 2009; e.g. Autran and Peterlongo, 1973; Lotout, 2017).

Two other key examples of eclogite-facies metagranitoids containing jadeite belong to the Alpine orogen. These are the Monte Mucrone granitoid in the Sesia Unit (HP event between 500 and 625 ∘C and 1.3–2.5 GPa: Roda et al., 2012) and the Brossasco–Isasca granitoid in the Dora-Maira Massif (HP event at 730 ∘C and 4.0–4.3 GPa: Ferrando et al., 2009), both located in the Western Alps. In both cases, the HP overprint is revealed, as in La Picherais metagranite, by pseudomorphs after plagioclase and various coronae at the grain boundaries between magmatic minerals. Similarly to all other examples, the igneous biotite at contact with plagioclase is partly replaced by phengite and rimmed with a garnet corona that contains a few rutile inclusions on the biotite side and is intergrown with quartz on the feldspar side; this garnet also shows a sharp increase in grossular content towards plagioclase (Biino and Compagnoni, 1992; Rubbo et al., 1999; Tropper et al., 1999; Bruno and Rubbo, 2006). The biotite–K-feldspar coronae are similar to the Essarts coronitic gneisses with a succession of garnet / garnet + quartz / phengite + quartz without albite (contrary to La Picherais metagranite). A major difference of the Monte Mucrone occurrence with the Variscan examples is the presence of a symplectitic rim, in the biotite–plagioclase coronae, composed of phengite + quartz ± jadeite or garnet + jadeite ± phengite ± quartz towards plagioclase. This symplectite is substituted by an albitic rim in the Variscan occurrences. A second difference is that the magmatic plagioclase was pseudomorphed into jadeite-bearing assemblages, typically jadeite + zoisite + K-feldspar + quartz ± kyanite ± phengite at Monte Mucrone (Rubbo et al., 1999; Tropper et al., 1999) and jadeite + zoisite + K-feldspar + quartz + kyanite in the Dora-Maira Massif (Biino and Compagnoni, 1992; Bruno et al., 2001). These major differences could be explained by equilibrium issues, due to kinetic factors (e.g. amount of fluids, residence time), the La Picherais metagranite having been metastable at peak conditions, or by a greater retrogression leading to the full disappearance of jadeite ± kyanite.

As shown in the previous section, sodic clinopyroxene is generally missing in poorly deformed metagranitoids despite being expected from phase equilibrium calculations. An example is the Tso Morari metagranite which lacks jadeite relic despite reaching the coesite stability field (Bidgood et al., 2023). This is also the case in many HP orthogneisses where sodic clinopyroxene is rarely found (e.g. Young and Kylander-Clark, 2015; Palin et al., 2017). A key example is the Western Gneiss Region (WGR) in western Norway where felsic rocks do not contain clinopyroxene except for isolated cases (e.g. Engvik et al., 2000; Krabbendam et al., 2000; Wain et al., 2001) and rarely contain kyanite despite peak conditions higher than 2.5 GPa (Young et al., 2007; Young and Kylander-Clark, 2015). Therefore, we consider three different scenarios for the La Picherais metagranite, whose HP coronae and pseudomorphs would have formed (1) at peak conditions, with jadeite ± kyanite lately replaced during retrogression; (2) during cooling and exhumation; and (3) during prograde metamorphism, having remained metastable at peak conditions.

To support the first hypothesis, we need clear textural evidence of sodic pyroxene breakdown. However, no pyroxene relic, no Amp + Pl symplectite or pseudomorph after pyroxene was found. The honeycomb-like structure described by Godard (2009) was not observed either. This situation is similar to what Young and Kylander-Clark (2015) described in the WGR (Norway), where sodic pyroxene was not preserved in thousands of square kilometres of HP quartzofeldspathic rocks. In HP orthogneisses, the non-conservation of jadeitic pyroxene could be explained by deformation, hydration and retrogression during exhumation, resulting in the formation of retrograde Pl + Bt salt-and-pepper textures replacing HP parageneses (e.g. Hacker et al., 2010). However, in the case of the undeformed La Picherais metagranite, the HP coronae are perfectly preserved, with only minor retrograde biotite replacing phengite. In addition, the conspicuous presence of albite in the coronae at the biotite–K-feldspar interface suggests the mobility of Na2O that was not fixed by the crystallization of clinopyroxene. It is therefore difficult to validate the first hypothesis.

The second hypothesis is based on the observation of granulites (White and Powell, 2011; Doukkari et al., 2018; Schorn et al., 2020) and mafic eclogites (Baldwin et al., 2015; Li et al., 2016; Vrabec et al., 2012) in which coronae and pseudomorphs are mainly interpreted as retrograde features formed during slow cooling and exhumation of high-grade metamorphic rocks. The corona textures would have formed because of the high chemical potential between the initial/prograde minerals. At La Picherais, the formation of the coronae and pseudomorphs at the onset of the retrograde path is supported by the composition of some HP minerals, in particular the Si content of phengite in the biotite–plagioclase coronae and the relatively anorthite-rich neo-plagioclase in the pseudomorphs after plagioclase and ilmenite–plagioclase coronae. However, purely retrograde coronae and pseudomorphs imply that no reaction occurred during prograde metamorphism, which could only be explained by fast burial and lack of water. Without evidence for the influx of a retrograde fluid, it appears that most of the water was released during the partial breakdown of biotite (Schorn, 2022), which is typically a prograde process.

The third hypothesis implies a prograde incipient equilibration, predominantly occurring in the stability field of albite, followed by the metastability of the latter in the stability field of jadeite. This hypothesis was suggested for a large number of HP felsic rocks (e.g. Gil Ibarguchi, 1995; Rubbo et al., 1999; Bruno et al., 2001; Young and Kylander-Clark, 2015; Schorn, 2022) and would result from the sluggish kinetics of jadeite crystallization (Young and Kylander-Clark, 2015). A number of kinetic factors such as the availability of fluid, bulk composition and reactivity, and residence time at depth may impede equilibration at high pressures and may prevent the formation of jadeitic clinopyroxene (Heinrich, 1982; Austrheim, 1987; Rubie, 1990; Rumble, 1998; Guiraud et al., 2001; Proyer, 2003; Holyoke and Tullis, 2006; Zhao et al., 2006; Peterman et al., 2009; Spencer et al., 2013; Young and Kylander-Clark, 2015; Schorn, 2022). Nucleation barriers are yet a key parameter for other HP minerals that preferentially crystallized along intragranular cracks or grain boundaries (forming coronae), and nucleation is even more difficult for clinopyroxenes, whose crystallization requires a significant P–T deviation from the producing reaction, or a structurally similar template mineral (e.g. pre-existing clinopyroxene; Wain et al., 2001; Bras et al., 2021), and is facilitated by significant presence of fluid (Rubie, 1998; Schorn, 2022). Schorn (2022) proposed that the H2O produced by partial biotite breakdown in metagranitoids is sufficient enough to form HP coronae and pseudomorphs. Nonetheless, he explains that the amount of water needed to reach saturation increases with pressure and its limited availability, in a closed system, may impede jadeite crystallization (Young and Kylander-Clark, 2015; Schorn, 2022).

We suggest that the coronae formed during prograde metamorphism remained metastable in the jadeite stability field because of nucleation barriers and the decreasing availability of water with pressure. The corona minerals, mostly the albite, may have also partly re-equilibrated during retrogression when the water availability increased again.