Assessing Chemical and Mineralogical Properties of the Alpine Slab Based on Field Analogs and Ambient Noise Tomography

Recent geophysical campaigns in the Alps produce images with seismic property variations along the slab of sufficiently fine resolution to be interpreted as rock transformations. Since the reacting European lower crust is presumed responsible for the variations of velocities at the top of the Alpine slab, we sampled local analogs of the lower crustal lithologies in the field and modeled the evolution of equilibrium seismic properties during burial, along possible pressure‐temperature paths for the crustal portion of the slab. The results are then compared to the range of the S‐wave velocities obtained from the S‐wave velocity tomography model along the CIFALPS transect. The velocity increase from 25 to 45 km within the slab, in the tomographic model is best reproduced by the transformation of specific lithologies in the high‐pressure granulite facies along a collisional gradient (30°C/km). Although the crust is certainly not completely homogeneous, the best candidates for the rocks that make up the top of the Alpine dip crustal panel are a kinzigite from Monte San Petrone, a gneiss from the Insubric line, and blueschist mylonite from Canavese. While they may not represent the entirety of the crust, they are sufficient to explain the tomographic velocity of the Alpine slab. A lateral lithological contrast inherited from the Variscan orogeny is not required. Eclogitization, suggested as the first‐order transformation in convergence zones, could be a second‐order transformation in collisional wedges. These results also imply a partially re‐equilibrated thermal gradient, consistent with the Alpine thermal state data at depth.

gravimetric data reveals that eclogitization of the Indian LC is delayed until the first dehydration reactions of the underlying mantle occur (Hetényi et al., 2021).These interpretations are not limited to large active convergent plate boundaries.In the Alps, current studies (Hetényi et al., 2018;Paul et al., 2021;Zhao et al., 2015) provide geophysical images of sufficiently high resolution to allow this same effort of interpretation (25 × 25 × 15 km for the V p crustal structure, Diehl et al. (2009), or up to 15 × 15 × ∼10 km, Solarino et al., 2018).In the Alps, the metamorphic transitions (e.g., blueschist to eclogite) are likely to be imaged by tomography because they occur as sharp to tens of kilometer-wide boundaries (e.g., the LT to the HT blueschist-facies transition in the Schistes Lustrés (SL) unit, Herviou et al., 2022).
The interpretation of these images and the determination of the rocks composing the crustal top of the Alpine panel require independent knowledge of the seismic velocities of the rocks potentially involved.Several worldwide databases of crustal rock seismic properties exist in the literature (e.g., Barruol & Mainprice, 1993;Lloyd et al., 2011;Rudnick & Fountain, 1995;Rudnick & Gao, 2003) and were recently used in the Alps (Malusà et al., 2021).However, the exact position of metamorphic transitions and associated velocity contrasts strongly depends on the chemical composition of rocks (e.g., Almqvist & Mainprice, 2017).The use of generic databases is hence an oversimplification that only gives first-order insights.Therefore, we need studies based on local analogs to refine our interpretations.To this end, it is crucial to build a regional catalog of quantified seismic properties, considering chemistry, pressure, temperature and metamorphic equilibration.The existence of such a catalog would be a precious tool to detect and identify the reaction fronts and infer the most probable lithologies constituting the crustal top of the Alpine slab at depth.Initial interpretations of the NRP-20 WEST and ECORS CROP profiles proposed a duplex of LC beneath the internal Alps (Schmid & Kissling, 2000).The reinterpretation of the same transects in the light of new tomography results (Diehl et al., 2009), did not confirm such a duplex (Schmid et al., 2017).Thus, although this idea is not new (Marchant & Stampfli, 1997), the most recent interpretations consider that the European LC does not significantly participate in the crustal wedge formation and must therefore remain part of the down-going slab.Furthermore, balanced cross-sections through the Western Alps indicate a shortening of up to 66 km (Bellahsen et al., 2014).This shortening is consistent with a collision wedge composed mostly of upper crust.Finally, there is no evidence of a continuous unit derived from the European LC at the surface (Schmid et al., 2004), such as the Sesia Zone (SZ) for the Adriatic plate.Hence, the European LC is probably responsible for the strong visible wave conversions along the slab (e.g., Nouibat et al., 2022;Zhao et al., 2015).The presence of a large body of serpentinites in the subduction channel, metasomatized mantle peridotites and eclogitized LC have also been suggested in the interpretation of CIFALPS seismic images (Malusà et al., 2021;Zhao et al., 2020).A way to elaborate on this hypothesis of a LC component at the top of the subducted slab is to find disseminated outcrops of analogs of the LC in the field and model the evolution of their seismic properties during burial and metamorphic equilibrations."Lower crust" is here understood in the sense of "lower crustal level" and not in the geochemical sense as in most global reviews of crustal composition (Lloyd et al., 2011;Rudnick & Fountain, 1995;Rudnick & Gao, 2003).
This study aims to critically assess the Alpine rocks as local analogs of crustal units of the subducted Alpine panel.The evolution of seismic properties of equilibrium rocks during burial, that is, along possible P-T trajectories for the slab, is predictably based on the evolution of its mineralogical assemblage and the thermoelastic parameters of each mineral (e.g., Almqvist & Mainprice, 2017).This provides the evolution of the isotropic component of the velocity tensor with depth.The obtained results are then compared to the seismic S-wave velocities from the tomographic model of Nouibat et al. (2022), along the CIFALPS profile, for which not only the V s values but also their uncertainties are available.We discuss the results with emphasis on the interpretation of the velocity increase before 50 km depth within the crustal top of the slab as due to the presence of felsic to intermediate subject to progressive granulite-facies imprint.The implications for the evolution of the Alpine thermal state since the end of subduction are also discussed.

Seismic Velocities of the Alpine Slab Along the CIFALPS Transect
Along the CIFALPS transects, imaged at a resolution of 7 × 5-10 km in the horizontal direction and 0.2 km in the vertical direction (Paul et al., 2022;Zhao et al., 2015), the top of the European dipping slab appears connected to the European LC in P-wave to S-wave conversion models.Along the same transect, tomographic models, whose data are independent of the receiver functions, also show the same, in (a) P-wave tomography modeling, with a horizontal resolution of 20 km and a vertical resolution of 2-3 km (Solarino et al., 2018), and in (b) S-wave tomography model from ambient noise, with a horizontal resolution of 15 km and a vertical resolution of 1-4 km in the crust to 5-10 km in the mantle (Nouibat et al., 2022).Moreover, for the first time, this latter S-wave tomography model provides not only the velocities but also the uncertainties (from 0.15 to 0.20 km/s) associated with the group velocities, allowing critical geological interpretations.It is indeed possible to assess whether velocity variations resulting from rock transformations during burial lay within the model uncertainty or not.For this purpose, in the S-wave ambient noise tomography model (Figure 1; Nouibat et al., 2022), S-wave velocity curves as a function of depth are extracted along the CIFALPS transect from 2 areas: (area A) at the front of the mountain belt and in the foreland (Figures 1 and 2), as a representative of the European plate, and (area B) within the mountain belt, along the plunging slab defined based on receiver-function analyses (Figures 1 and 2).
The profile of the European plate corresponds to the data 15 km from the western end of the transect.Consideration of the probabilistic uncertainty of the S-wave tomographic model for each pixel allows to obtain ranges of possible seismic velocities for different domains across the chain.The uncertainty profile (Figure 1c) shows that the resolution is highest at velocity contrasts, such as the European Moho, which correspond to a standard error close to 2% (Figure 1c).Within the dipping panel, the velocity contrasts shallower than 50 km are better constrained than those deeper (4% vs. 6 or 7%, Figure 1c).At mantle depths (i.e., using the geophysical Moho as 4.3 km/s velocity contour), European seismic velocities are 4.4 km/s ± 0.2 km/s (Figure 2).At crustal depths, the velocity profile reaches 3.8 km/s at 20 km depth (Figure 2).Along the plunging slab, seismic velocities are lower than those in the European LC at the front of the mountain belt, that is, mostly lower than 4.0 km/s (Figure 2).At the same depth, the velocity therefore decreases from the front of the belt to the slab.Since the slab dipping is continuous (as shown in the receiver function model of Zhao et al., 2015; Figure 1a), this  decrease is evidence of probable crustal transformations (deformation and/or metamorphism of the crust).From 20 to 50 km depth, seismic velocity increases rapidly from 3.3 to 3.9 km/s, which is beyond the resolution within the dipping slab (±0.2 km/s, Figures 1c and 2).A decrease is computed between 50 and 65 km depth, and then a second increase from 65 to 80 km appears but with a lesser confidence value (±0.4 km/s, Figures 1c and 2).
In this prospect, the European slab is finally characterized by two successive velocity steps: a well-defined first one at a shallower depth (25-45 km) and possibly a deeper and less marked one.The present study mainly focuses on the significance of the shallowest and best defined velocity step.

Sampling Strategy
Assuming a crustal part of the slab made of lower crustal rocks, and searching outcropping analogs, the rare field occurrences of the European LC can be classified into three different rock types.First, we sampled rocks in the external zone considered as unmodified or slightly modified basement (i.e., with no record of Alpine subduction, and therefore close to the post-variscan protolith of the European margin basement).We also sampled rocks from the internal zone, which have undergone significant burial.They represent buried metastable or reacted continental crust.Finally, we sampled rocks from the Adriatic plate, as mirror analogs of the European margin since the Adriatic LC can be considered as a remnant of the conjugate margin of the European crust (Manzotti et al., 2014).In addition to field analogs of the LC, we also sampled the rocks juxtaposed to them along supposedly deep contacts.

The External Crystalline Massif
The External Crystalline Massifs (ECM, Figure 3) constitute the upper to mid-crustal level of the ancient European margin (e.g., Bellahsen et al., 2012;Bellanger et al., 2015).Within these massifs, the metric-scale interbedded mafic rocks recorded high-pressure metamorphism in the Middle Carboniferous (Fréville et al., 2022).These  Bellahsen et al., 2014).They constitute good analogs for the unmodified basement of the pre-Alpine European margin.The mafic intercalations sampled (Figure 3) include amphibolites and retrogressed eclogites expected to be also present in the LC.

The Gruf Complex
The Gruf complex is located south-east of the Lepontine Dome (Figure 3) and north of the Insubric Line (IL).It consists of migmatitic orthogneiss (including granulites), paragneiss, mica schists and associated leucogranite (Galli et al., 2011;Mintrone et al., 2022).The age of their ultra-high temperature peak is debated and could be Permian, Alpine, or both (Galli et al., 2012;Nicollet et al., 2018;Schmitz et al., 2009).These rocks constitute anyhow relics of the LC.Representative biotite orthogneiss, deformed leucogranite and charnockite of the Gruf Complex have been sampled.

The Adriatic Lower Crust, Ivrea Zone
The Ivrea-Verbano Zone (IVZ, Figure 3) preserves a continuous section through the Adriatic Permian lower continental crust (e.g., Boriani et al., 1990;Handy et al., 1999;Zingg et al., 1990), separated from the accreted distal Adriatic margin and Penninic domains of the Western and Central Alps by the IL.The IVZ is subdivided into two major units, the Kinzigite Formation and the Mafic Complex.The first, dominated by amphibolite-to granulite-facies metapelites (e.g., Quick et al., 2003;Redler et al., 2013), was metamorphosed during Variscan orogeny (316 ± 3 Ma, Ewing et al., 2013).The second, dominated by gabbroic, and dioritic intrusive complexes, generated a contact metamorphism affecting only the first kilometers around them (e.g., Redler et al., 2012).Sampling was carried out along the Val Sesia and Val Mastallone in both the Mafic Complex and the Kinzigite Formation.We collected metabasite, garnet-bearing metagabbro, granite, and mylonitic paragneiss.

Rocca Canavese Zone
Within the SZ, the Rocca Canavese Thrust Sheet (RCT, Figure 3) is a distinct metamorphic unit located along the southern edge of the IL.It is characterized by a mélange of mantle and crustal lithologies (metapelites, metagranitoids, metabasics, and serpentinites, Roda et al., 2020).The block-in-matrix arrangement, lithological mixing, and contrasted PT-time paths suggest that mixing occurred at LP during subduction (Roda et al., 2018).The sequence sampled along the Lanzo-Canavese contact includes serpentinites, blueschist facies mylonites and biotite gneiss (Figure 3).We also sampled serpentinites as such lithology has been proposed as a component of the Alpine subduction channel (e.g., Malusà et al., 2021;Zhao et al., 2020).

Gran Paradiso-Schistes Lustrés Contact
Rocks along the Gran Paradiso (GP)-SL boundary were sampled as representative of the top contact of a continental crust panel.Paragneiss, orthogneiss and internal Briançonnais units make up the GP (e.g., Le Bayon & Ballevre, 2006) and form a tectonic window within the SL.The latter are composed of oceanic crustal fragments (metagabbros and metabasalts) and mantle rocks associated with marine metasediments of Jurassic to Cretaceous age (Deville, 1987).The continental units of the GP and the lower part of the SL were buried down to the eclogite facies (2.0 GPa and 500°C for the GP and 2.2 ± 0.2 GPa, 480 ± 30°C for the SL; Le Bayon & Ballevre, 2006;Plunder et al., 2012).In order to assess the signature of such a lithological contrast in seismic imaging, we sampled metabasites and mica schist from the SL and mylonitic gneiss and chlorite-bearing schist from the GP.

From Bulk Chemistry to PT Velocity Map
One way to predict seismic velocity at lower crustal conditions is to consider natural rocks as isotropic and to calculate their seismic properties from the relative abundance of mineral phases using their acknowledged properties (Abers & Hacker, 2016).With single-crystal elastic parameters, the average velocities of an aggregate can be calculated for given P-T conditions using the Bina and Helffrich (1992) formalism (Hacker et al., 2003).Thus, in order to predict the evolution of the seismic properties of the rocks at equilibrium along the burial path, the bulk chemical compositions of the rocks were determined and the equilibrium paragenesis was modeled along theoretical PT trajectories for a dipping panel.
In order to test the impact of the variation in the number of hydroxyl bearing minerals on the seismic velocities, the saturation state of the samples was calculated over a temperature range (500-800°C at 1 GPa), using the same solution models as above, and compared to the loss on ignition (LOI) measured.Corsican kinzigite exhibits two seismic velocity jumps resulting from metamorphic reactions (Figure 5c).As explained below, one occurs along the collisional gradient, associated with the amphibolites-to-granulites transition, while the other takes place along the subduction gradient, related to eclogitization.Therefore, due to the significant potential for the reaction of this rock, this sample was chosen to assess the impact of rock hydration on seismic velocities.These were calculated in three states: undersaturated, undersaturated at lower temperatures (i.e., below 600°C) and saturated.

Seismic Properties Modeling Within a PT Domain
Rock seismic velocities (V p , V s , and V p /Vs) within a pressure-temperature domain were calculated from single-crystal physical properties (database from Abers & Hacker, 2016).The rock mode was translated into volume percentage and mineralogical phases representing less than 1% were removed.The finite Eulerian strain was calculated from the pressure and the material parameters by assuming the third-order finite strain equation using the Bina and Helffrich (1992) equation.Thermal expansivity, density, bulk and shear modulus were then extrapolated over a pressure range from 0.25 to 2.5 GPa and a temperature range from 100°C to 1,000°C, allowing the eventual estimates of V p and V s .
Phases for which elastic parameters are not known were substituted by close end-members with available parameters (see Supporting Information S1 for more details).The Voigt-Reuss-Hill (VRH) arithmetic mean has been used to calculate seismic properties.

Seismic Properties Along PT Profiles
In this study, rock re-equilibration was modeled according to several end-member scenarios: (a) a typical Alpine subduction gradient.A constant 8°C/km gradient was chosen for representing the subduction-related PT profile based on the metamorphic records from exhumed HP-LT units in the Western Alps (Agard, 2021) and (b) a collision-related thermal gradient.We used a second order solution of the heat equation fixing surface temperature and geotherm at 25°C and 30°C/km, respectively, and the lithosphere asthenosphere boundary at a temperature of 1,300°C with an adiabatic geotherm at 0.4°C/km; (c) a collision-related thermal gradient, along which the rocks did not react and have kept a paragenesis typical of Alpine subduction.This latest case is explored to discriminate between temperature effects and mineralogy effects.PT profile calculations were performed according to the previous scenarios.For conversion of pressure to depth, we only considered lithostatic pressure along the slab and we calculated it as follows: with densities from the CIFALPS density model previously used for fitting gravity data (Zhao et al., 2015).Melts are produced in 5 lithologies (0.5 wt% in the Canavese mylonitic blueschist, 0.16 wt% in the Val Sesia metagabbro, 0.08 wt% in the Taou Blanc chlorito-schist and 0.02 wt% in the Val Sesia gneiss) along the HT profile.Although we acknowledge that the presence of melt could have a tremendous impact on effective wave velocities (Hacker et al., 2014), the present calculations are solely based on the solid rock part.The absence of present-day volcanism, the limited heat flow in the western Alps, and the age of their youngest magmatic intrusions (26-24 Ma for Novate granite, Lustrino et al., 2011) indeed point to the absence of a significant melt fraction at depth today.Moreover, a significant melt fraction would result in a significant drop in the seismic velocity, which is not observed in the profiles.

Sample Description
A short microscopic description of the samples is given in Table 1, accompanied by microphotographs of thin sections (Figure S4 in Supporting Information S1).

Bulk Chemical Composition of Rocks and Choice of Representative Analysis
The chemical compositions of major elements were measured for 48 sampled rocks (Figure 4 and Table S1 in Supporting Information S1).The felsic  Whitney and Evans (2010).

Table 1
List of Samples for Which S-Wave Velocity Has Been Calculated, Classified According to the Topology of the S-Wave Velocity Maps composition range includes 14 samples, 5 are in the intermediate one, 22 have a mafic composition and 7 are ultramafics.Out of the 48 samples measured, 8 were chosen as representative for the chemistry range explored to compute the evolution of their seismic velocities during burial and heating.
On the alkalis sum diagram, the samples are distributed over a range from 40% to 85% SiO 2 and 0%-9% alkalis (Figure 4).Almost half of the samples lay in the basalt and its adjacent domains (trachy-basalt, tephrite and basalt andesite) with less than 6% alkalis, while the others are clustered in the dacite to rhyolite domain.Mafic rocks have a Mg# higher than 0.4, while that of felsic rocks is always below 0.4 (Figure 4).The LOI varies from 0% to 5% with a majority below 4% (Figure 4).
Of these 48 rocks, 8 samples representative of the chemical diversity were selected for velocity calculations.From the Canavese, we selected a mylonitic blueschist (4ALZ20); from the Gruf complex, a charnockite (11GR21); from GP-SL contact, a garnet-bearing gneiss (7BGP20) and from the Taou Blanc massif a chlorito-schist  (29GP20); from Val Sesia, a garnet-bearing metagabbro (2AIV20) and a gneiss of the IL (7IV20); from Monte San Petrone Unit (Corsica), a kinzigite (3COR21); from the Argentera series (ECM), an amphibolite (AN1907).According to the topology of the obtained S-wave velocity maps, the samples can be classified into 3 groups (Figure 5).Indeed, instead of relying only on chemical classification, rocks have been grouped based on the visible seismic velocity fields within the studied PT range.While the rock chemical composition influences seismic field mapping by governing mineralogical assemblages, these groups are not chemically homogeneous.For instance, as shown below, group 1 is composed of two felsic samples and one ultra-mafic sample.What unites them is their limited potential for reactions.Therefore, while it is related to its chemical composition, the rock's capacity to react, "its fertility", appears to be a more relevant factor to consider.The first group includes samples for which seismic velocities are mainly governed by the plagioclase breakdown to pyroxene reaction (Figure 5a), which is known to induce a sharp impedance jump in the continental crust (Diaferia & Cammarano, 2017).The second group corresponds to samples whose velocity maps highlight the transition from amphibolites-to the HP granulites-facies (Figure 5b).The last group is made of samples whose velocity maps emphasizes the blueschist to the eclogite facies transition in addition to the amphibolites-to the HP-granulites facies transition (Figure 5c).

Evolution of S-wave
The first group governed by the plagioclase out joint includes 3 samples: the GP gneiss, the Gruf charnockite and the Val Sesia metagabbro.The Ab = Jd + Qz boundary limits two domains with V s = 3.50 or 3.60 km/s and V s = 4.00 or 4.50 km/s (Figure 5a).This boundary corresponds to a change from a plagioclase-dominated paragenesis with minor quartz at low pressure to quartz-jadeite dominated association at high pressure for the GP gneiss and the Gruf charnockite.Garnet and kyanite in a significant proportion are only computed for the Val Sesia metagabbro (Figure 5a).Thus, although their velocity topology is the same, velocity ranges differ from low values (from 3.6 to 4 km/s) for the most felsic lithology (the Gruf charnockite) to higher values (from 4 to 4.5 km/s) for the most mafic lithology (the Val Sesia metagabbro).This trend is not new.Lloyd et al. (2011) states: "According to recent compilations (e.g., Rudnick & Fountain, 1995;Rudnick & Gao, 2003), increasing average P-wave seismic velocities with depth indicates increasing proportions of mafic lithologies."This paper also explains how the presence of felsic rocks is necessary to match the P-wave velocities of the average middle crust, as a mixture of mafic rocks would generate excessively high velocities (e.g., Rudnick & Fountain, 1995;Rudnick & Gao, 2003).
The second group, governed by the amphibolites-to granulites-facies transition, consists of 3 samples: the Canavese mylonitic blueschist, the Val Sesia gneiss and the Argentera amphibolite.HP granulites-and amphibolites-facies and the Jd + Qz field are systematically delineated in the seismic velocity map of these samples (Figure 5b).The transition from amphibolites-to HP granulite facies is manifested by a mean velocity jump of 0.40 km/s (mainly from 3.60 or 3.80 km/s to 4.20 or 4.40 km/s, Figure 5b), explained by the changes from low nominal velocity minerals (plagioclase, amphibole and mica, Abers & Hacker, 2016) at LP/HT, to high nominal velocity minerals (garnet, kyanite, and pyroxene, Abers & Hacker, 2016) at HP/HT (Figure 5b).
The last group exhibits both the amphibolites-to granulites-facies and blueschist-to eclogite-facies transitions.It consists of two samples: the Taou Blanc chlorito-schist and the Corsica kinzigite (Figure 5c).The amphibolitesto granulites-facies transition appears similar to group 2 as expected.The eclogitization is shown by a velocity increase of 0.3 km/s for the kinzigite and 0.1 km/s for the chlorito-schist at 65 km depth.The velocity increase in the last sample is very small for eclogitization as along this cold subduction gradient, the eclogite facies is not completely reached.According to the seismic velocity map, a full eclogitization would generate a velocity variation of around 0.4 km/s.This increase is explained by the growth of high nominal velocity minerals (such as garnet, pyroxene, Abers & Hacker, 2016), even though low nominal velocity minerals appears (lawsonite e.g., Abers & Hacker, 2016).

S-Wave Velocity Along Subduction or Collision PT Profile
Samples from group 2 (i.e., those highlighting only the amphibolites-to granulites-facies transition) show nearly constant velocities along the subduction gradient, reflecting minor paragenesis changes as the high-velocity domain within the eclogite facies is not reached (Figure 5b), while samples from group 3, reaching the eclogite-facies, show a step of 0.20 km/s at 65 km depth (Figure 5c).Along the collisional gradient for both groups, a significant velocity increase (around 0.2-0.3km/s) is consistently observed at comparable depths: 3.60-3.80km/s at 35-45 km depth for the Canavese mylonitic blueschist, 3.70-3.90km/s at 32-38 km for Val Sesia gneiss and the Argentera amphibolite for the group 2, and to 3.80-4.10km/s at 25-32 km for the Corsica kinzigite and 3.90-4.10km/s at 33-40 km for the Taou Blanc chlorito-schist respectively for the group 3 (Figures 5b and 5c).

Impact of Water Content on S-wave Velocity
In order to assess the impact of the hydration state of the rock on the calculated seismic velocities and since present-day LOI does not represent the hydration state of rocks at lower crustal depths, their saturation curves were calculated on a temperature range from 500 to 800°C (Figure 6).Knowing the saturation state of the rock, hydrated or dehydrated scenarios can be explored.Two situations can be considered: either rock contain more water at depth than they do after exhumation or they may have gained water during exhumation.The different hydration states considered (i.e., saturated, highly or slightly undersaturated) allow exploration of situations expected at lower crustal depths in a quasi-static system for 5 Ma (i.e., the end of the collision, Bellahsen et al., 2014).Depending on the initial bulk rock H 2 O content, lithologies will reach their saturation curve (Figure 6) at different temperatures and hence different depths along the profile.
The stability of free water in the system was calculated for each quantity of H2O evaluated in the PT range studied.All stability limits were then plotted together to allow comparison between behaviors.The saturation thresholds vary from 0.5 to 3 wt% H 2 O.The Gruf charnockite and the GP gneiss show almost constant H 2 O% for saturation over the temperature range studied (Figure 6), while the Val Sesia metagabbro displays a higher variation from 3 wt% at 500°C to 0.5 wt% at 800°C (Figure 6).For the other samples, their H 2 O % for saturation varies from about 2.2 wt% at 500°C to 1 wt% at 800°C (Figure 6).We chose the Corsica kinzigite as an example to evaluate the impact of H 2 O content on seismic velocity as it belongs to group 3 and presents the two velocity jumps linked to the HP-granulites and eclogites facies boundaries (Figure 5c).Along the subduction gradient, a velocity increase of 0.4 km/s at 65 km depth due to eclogitization is mostly visible for the saturated state (3 wt% H 2 O, Figure 6).The under-saturated evolution at lower temperature (1.5 wt% at 500°C) shows a lower velocity increase (0.1 km/s, Figure 6) some kilometers deeper.Along the collision gradient, the velocity jump is also visible in the saturated and the undersaturated evolutions at LT and its amplitude is almost 0.3 km/s at 25 km depth (Figure 6).Therefore, the velocity jumps are present only for rocks above or close to their water saturation.The amplitude of this velocity jumps increases with water saturation.

Comparison Between Seismic Velocities From Geophysical Data and Calculated S-wave Velocities
S-wave velocities calculated from tested lithologies are compared to the range of those along the Alpine slab, as imaged along the CIFALPS transect.

A Poorly Reacted Felsic to Intermediate Crustal Slab Down to 70 km
Along the subduction thermal gradient, all mafic to intermediate samples yield much higher velocities than those computed along the slab (4.65 km/s for the Val Sesia metagabbro, 4.55 km/s for Taou Blanc chlorito-schist, 4.35 km/s for the Corsica kinzigite, and 4.10 km/s for Canavese mylonitic blueschist; vs. 4.00 km/s maximum in the slab, Figure 5).These velocities are closer to those of the mantle at the front of the belt (from 4.00 to 4.90 km/s, Figure 5).Along the collision thermal gradient, the S-wave velocities of the samples are still outside the range of those of the Alpine slab, except for the Canavese mylonitic blueschist (Figure 5b).These calculations support the idea that a fully eclogitized mafic LC panel would have the same seismic properties as the surrounding mantle (4.4 km/s at 60 km) and hence would not be distinguished from the mantle in V s models (e.g., Christensen, 1989;Hacker & Abers, 2004;Hacker et al., 2003Hacker et al., , 2015;;Hetényi et al., 2007;Malusà et al., 2021).
By contrast, velocities of felsic samples are in the range of those of the Alpine slab: around 4.00 km/s and 3.60 km/s along the subduction and collision gradients for the Gruf charnockite and the Val Sesia gneiss (Figure 5).According to the V s model, the crustal panel would have a mainly felsic to intermediate composition with only a minor or highly hydrated mafic component, as any substantial interlayering of mafic lithologies would increase the effective velocity averaged by body wave propagation (e.g., Almqvist & Mainprice, 2017;Hacker et al., 2015;Rudnick & Fountain, 1995).Of course, this conclusion highly relies on the assumption that computing rock velocities from phase proportions and absolute phase properties actually captures the natural complexity.

Shallow Velocity Step: The Seismic Signature of a Granulites Facies Continental Crust?
The hypothesis of an eclogitized continental crust slab top (Figure 7, model 1.1 and preserved subduction thermicity sketch) to explain the velocity increase at 25-45 km depth, implies that the slab burial path crosses the eclogite-facies boundary at low temperature.This could lead to a substantial velocity jump only for lithologies governed by the plagioclase-out reaction, that is, the first group.Since the LC was certainly already at 500-600°C before its involvement in the wedge, this hypothesis also implies an isothermal burial of the rocks (Figure 7 model 1.1).This also implies the preservation of the 8°C/km thermal gradient along the plate boundary during the involvement of the proximal European continental margin within the convergent zone and hence long after the transition from subduction and the involvement of the Briançonnais distal margin.
An alternative to the eclogitized slab top hypothesis would be a burial along a gradient modified by the introduction of the continental margin in the convergent zone (Figure 7, model 2).The end-member of this hypothesis is the burial of the continental slab along a fully reequilibrated gradient, that is, along a collision thermal gradient (Figure 7, collision related thermal relaxation sketch).In this case, lithologies belonging to the second and third groups are relevant, as they show a systematic velocity increase close to the expected depth range (Figures 5b  and 5c).Whatever the privileged lithology, this scenario implies a slab reaching a temperature close to 800°C at 45 km depth.Therefore, the two end-member scenarios mainly differ on the thermal state recorded by the continental slab at the time it was buried and since.Any intermediate scenario, with a background thermal regime warming while the continental slab is progressively involved in the subduction, would lead to an intermediate PT profile (Figure 7, model 3).

Seismic Signature of a Thermally Re-Equilibrated Slab
In scenarios 1.1 and 2 (Figure 7), calculations of seismic velocities consider total and instantaneous reactions at each PT condition.To test the possible effect of prograde equilibration during subduction followed by thermal re-equilibration without significant reactions (model 1.2, Figure 7), possibly due to the absence of fluids and/ or strain, modes were calculated along a subduction gradient, and then seismic properties were computed with temperatures along a collision gradient to reproduce a late heated metastable slab.This approach allows us to distinguish the effect of the thermal state and that of the mineralogy alone.The evolution of seismic velocities in the only thermally re-equilibrated scenario (small dotted line Figure 5) are close to 1.1 scenario (1.2, Figure 7).This means that the effect of thermal state alone is minor, compared to the effect of mineralogy.The difference is actually lower than the uncertainty on seismic velocities and hence the difference between scenarios 1.1 and 1.2 cannot be assessed with the present data set.Derivatives of S-wave velocity have been published for some analogous lithologies (Zertani…).For granulites, for example, an average ∂V/∂V = −2.45× 10 −4 km s −1 °C −1 and an average ∂V/∂P = 0.75 × 10 −4 km s −1 MPa have been assessed.By taking a ∆T = 400°C and a ∆P = 1 GPa, ∆V s = −0.023km/s, which is well below the uncertainty in topographic seismic velocities.

Role of Anisotropy in Seismic Velocity Calculations
The velocity step at 45 km depth in the S-wave tomographic model is located in the subducted slab of the European LC at a local increase in the slab dip (Figures 1 and 2).The bending of a seismically anisotropic slab is known to induce distortion in receiver functions (e.g., Schneider et al., 2013;Shiomi & Park, 2008).In a teleseismic P-wave tomography model, the illumination of structures is not identical in the three directions.The present data set is based on ambient noise shear wave velocity estimates.Thus, the structures are imaged from the shear wave component of horizontally propagating Rayleigh waves.Regarding this, it is reasonable to inquire whether the velocity increase, suggested as the transition from amphibolites to granulites in the rocks, could be attributed to the influence of rock anisotropy.However, predicting the effect of rock anisotropy on apparent seismic veloci ties according to the wave propagation method is not straightforward.Assuming possible configurations between the wave propagation direction, the mineral fast axis direction, the arrangement of this axis within the foliation plane, and the metric to kilometric layering of the crust, and deducing the consequences on apparent velocities in geophysical data, appears to be highly exploratory.
Compared to the mantle, the sources of crustal anisotropy are numerous.In addition to rock texture, banding and layering, the presence of fractures and fissures in the middle and LC (Siegesmund et al., 1991) and fluids (Almqvist & Mainprice, 2017) are possible causes of anisotropy.Interpreting seismic anisotropy as the result of ductile deformation in the crust, leading to a strongly preferred alignment of crystallographic axes and rock texture (Almqvist & Mainprice, 2017), is therefore only possible through quantified seismic properties.The preferred crystallographic orientation (crystallographic preferred orientations (CPO)) due to the rock fluage is recognized as the main factor influencing the direction of seismic wave propagation.Consequently, many studies focus on measuring the CPO of crustal minerals.As a result, the most anisotropic crustal minerals are amphiboles and mica (e.g., Lloyd et al., 2011;Naus-Thijssen et al., 2011).According to Christensen and Mooney (1995), velocity gradients appear to result from an increase in the degree of metamorphism and a decrease in SiO 2 %.
The anisotropy of felsic lithologies is therefore dominated by mica and can reach over 7% of P-wave or S-wave anisotropy (Barruol & Mainprice, 1993), although amphiboles can generate a certain degree of anisotropy (e.g., an AV p ranging from 9.4% to 12.5% for polycrystalline glaucophanes, Park & Jung, 2022 and AV p ∼ 38% and AV s ∼ 27% for a single crystal, Bezacier et al., 2010).Llana and Brown (2012) determined by calculation and direct measurements AV p ranging from 8.0% to 11.3% and AV s ranging from 7.9% to 10.5% in felsic gneisses.Among the felsic lithologies studied, two (the charnockite and the gt-bearing gneiss, Figure 4) are composed of a ranging from 5% to 7% of mica and from 0% to 7% of amphiboles (Figure 5a).They also include feldspars, quartz, and clinopyroxenes (Figure 5a), which are weakly anisotropic minerals (Almqvist & Mainprice, 2017).Therefore, the expected rock anisotropy is certainly not very high.In addition, the CPO of different minerals generate destructive seismic anisotropies resulting in relatively low values for the rock.It has been shown, for example, between quartz and mica (Lamarque et al., 2016;Ward et al., 2012) or between feldspar, quartz, and pyroxene (Lamarque et al., 2016).The third felsic sample in this study, the Sesia gneiss, is composed of 32% mica, which destabilize upon exiting the amphiboles-facies, and feldspars and quartz, whose proportions increase (Figure 5b).Unless the interplay of CPO erases the mica anisotropy, a significant anisotropy decrease is expected at the amphibolites-to-granulites transition.Erdman et al. (2013) calculated seismic properties for different metamorphic lithologies in the middle to LC of the Basin and Range.For all tested lithologies, they demonstrated that the anisotropy exhibits a slow velocity axis normal to the foliation and within the foliation plane, the properties are isotropic.Okaya et al. (2019) clarify that this conclusion cannot be generalized.If the rocks have undergone cylindrical folding, the produced anisotropy is gradually modified to the point where the distribution of structures becomes too heterogeneous for the assumption of transverse isotropy with a single axis of slow velocities to be valid in the crustal volumes.Similarly, despite the strong anisotropy conferred by planar schistosity, the variety of folds and fabric typical of phyllosilicate-rich rocks within larger-scale crustal volumes can lead to a decrease in anisotropy to the point where it may even appear isotropic (Naus-Thijssen et al., 2011).Since our samples do not exhibit cleavage or crenulation (Figure 4s), the hypothesis of a unique slow axis normal to the foliation and a decrease in anisotropy cannot be directly refuted.Thus, considering a case where the foliation plane containing the mica is horizontal and parallel to the direction of surface wave propagation, the resulting seismic recording is not affected by the vertical anisotropy of the mica.In this case, our calculated velocities are directly comparable to the ones derived from seismic data.Taking into account the effect of amphiboles, as the fast direction axis is contained in the foliation plane and parallel to the lineation, the logic is less obvious.
Eclogites exhibit an AV p < 5% and AV s < 4.5% (Llana-Fúnez & Brown, 2012) or less than 3% for both P-and S-wave anisotropy (Bascou et al., 2001).If their formation is associated with ductile deformation, AV p can reach 8% (Zertani et al., 2020).The lowest level of anisotropy is carried by the ultramafic rocks, with AV p ∼ 2-3% and AV s ∼ 0.8-1.5% (Llana-Fúnez & Brown, 2012).In eclogites and HP-granulites, the anisotropy is carried by pyroxenes (omphacite and diopside) and amphibole (Llana-Fúnez & Brown, 2012).The transition from amphiboles to granulites, which involves the destabilization of amphiboles in favor of garnets and pyroxenes, has also been quantified, resulting in a decrease in anisotropy (Brownlee et al., 2017).Among the tested lithologies, 2 samples are mafic, the Argentera amphibolite and the chlorito-schist (Figure 4).First, the amphibolites-granulites transition is expressed by a 32% loss of mica and a 5% increase in pyroxenes (Figure 5b).Although pyroxenes carry anisotropy like mica, their lower anisotropic capacity and percentage suggest a drop in anisotropy while an increase in velocity takes place.The second mafic sample, the chlorite-schist, displays a loss of 53% of amphiboles and a 38% increase in pyroxenes at the amphibolites-granulites transition (Figure 5c).Regarding the proportions of minerals involved, a lower anisotropy drop could be predicted.The velocity contrast of this sample associated with the amphibolites-granulites transition is consistent with that observed in the tomography model.A preserved anisotropy during this transformation, if it acts as a velocity decrease in the right direction, would align with those of the tomographic model.
Finally, our ultra-mafic sample, the Val Sesia metagabbro belonging to group 1, displays more than 40% of clinopyroxenes along the subduction gradient (Figure 5a) and 6% of amphiboles.Since pyroxenes are primarily responsible for the anisotropy of ultramafic rocks (Llana-Fúnez & Brown, 2012), it is possible that this sample exhibits a slight anisotropy.However, considering that the effect of anisotropy reduces the calculated seismic velocities for this sample in the right direction, its potentially weak percentage does not seem sufficient to explain the very low velocities of the Alpine slab in the tomography model.
As a result, most of the velocities calculated for the lithologies tested are slightly higher than those observed in the Alpine slab.However, the observed velocity jump is represented in the majority of lithologies and at consistent depths.In some lithologies (the Argentera amphibolite, the chlorito-schist and the Corsica kinzigite), this jump appears slightly too rapid.Since anisotropy is a factor not taken into account in our calculations, and since it allows us to lower these velocities, it is possible that these samples, containing amphibole, mica and pyroxene, are partially anisotropic.Direct measurements of seismic wave velocity and calculations of seismic properties from CPO measurements will help clarify the effect of anisotropy in our samples.

Implications for Present-Day Thermal State of the Alpine Slab
Seismic tomography in the Alps depends on the present-day thermal state of the collisional wedge, while the petrophysical properties of rocks at depth depend on the thermal state at the time they equilibrated.The heat flow at the surface, the seismicity within the prism, and metamorphic rocks at the surface provide some information on the past and current thermal state of the orogenic prism.Heat flow maps in the Alps (e.g., Pasquale et al., 2012Pasquale et al., , 2014;;Speranza et al., 2016) show values above 68 mW/m 2 at the western end of the CIFALPS profile, related to the West European Rift.In the subsiding Po plain, the heat flow drops down to 44 mW/m 2 .There is thus a strong regional lateral thermal gradient at the surface with the mountain belt itself being close to the acknowledged average for the continental lithosphere (60 mW/m 2 ).The mantle wedge seismicity has been used to provide indirect information on the thermal state of the slab, as some earthquakes are located at the depth of the S-wave model velocity step in the slab (∼50 km; Malusà et al., 2017).Malusà et al. (2017) inferred temperatures lower than 700°C for the mantle and hence a still cold continental slab, assuming that these earthquakes are related to the antigorite-out reaction, at 650-700°C for 2 GPa pressure (Fumagalli & Poli, 2005), in a cold and hence stiff mantle wedge.Earthquakes have nevertheless been documented at any depth in subduction mantle wedges including in hot, upper mantle beneath island arcs (L.T. White et al., 2019).Furthermore, the chlorite-out reaction occurring at 800°C for 1.5 GPa in lherzolites, that is, at 45 km depth along the slab, is a process also prone to generate earthquakes (Fumagalli & Poli, 2005).Using the same modeling approach as we used on the crustal analogs, we performed systematic calculations of V s for a serpentinite sampled next to the blueschist mylonite from Rocca Canavese-Lanzo contact (Figure S2 in Supporting Information S1).On this pseudosection, the chlorite-out reaction generates a higher velocity contrast than the antigorite-out reaction (respectively from 4.30 to 4.60 km/s to and from 4.20 to 4.30 km/s, Figure S2 in Supporting Information S1), which is more likely to be detected by the tomographic models.Thus, the occurrence of earthquakes in the mantle wedge is not univocal evidence of a cold wedge root beneath the Alps.Chlorite-bearing peridotite units in a hot environment could have the same behavior as serpentinites in a cold setting.
Finally, numerical models (Butler et al., 2013) used as evidence for a still cold Alpine slab do not integrate the end of convergence after collision.Using an average thermal diffusivity of 1.10 −6 m 2 s −1 (Gibert et al., 2003) for continental and mantle rocks at 700°C and using a scaling factor from the heat equation (  ∼ 2 √  ), thermal re-equilibration is predicted to be achieved in 7 Myrs for a 30 km thick cold slab anomaly (Syracuse et al., 2010).Timing of significant involvement of continental units in the Alpine convergent system can be bounded by the age of HP metamorphism in the most distal continental units and the mid-Eocene age of the last deposits in the Briançonnais (Barfety et al., 1992(Barfety et al., , 1995;;Jaillard, 1999), that is, 45 Ma in the Briançonnais units (Berger & Bousquet, 2008) and the age for mature collision at the Eocene-Oligocene boundary, based on the age of Barrovian metamorphic overprint in Lepontine Alps (35-28 Ma, Brouwer et al., 2005).Hence, the time span since the end of steady-state subduction thermal regime is anyhow greater than 20 Myrs.Whether the rocks now at 40 km depth only reacted when they were buried after 30 Ma or whether they also reacted during a recent thermal re-equilibration, it is highly unlikely that they developed and still preserve an HP-LT paragenesis.Eventually, magmatism documented along the IL between 33 and 30 Ma (e.g., Lustrino et al., 2011) evidenced lithospheric scale thermal relaxation at this time, possibly related to oceanic slab break-off (e.g., Rosenberg, 2004).
Therefore, these arguments do not support the idea of a present-day cold Alpine orogenic prism.Therefore, the slab velocity profile would have most probably been acquired along a partially reequilibrated thermal gradient (Figure 7, model 3).The shallow velocity step shown within the slab in the tomographic model would therefore highlight the entrance of felsic to intermediate rocks in HP granulite facies and the consumption of low seismic velocity minerals such as amphibole and/or biotite rather than the plagioclase-out reaction.This leads to the growth of mineral phases with higher seismic velocities, including garnet and pyroxene.

Possible Effects of Incomplete Reactions
As mentioned above, our calculations consider instantaneous and total reactions.Although the effects of kinetics have not been quantified in this study, it is possible to qualitatively predict its effect on seismic velocities.By considering transformations limited by reaction rates, the reaction fronts are expected to be shifted to greater depths along the slab (Hetényi et al., 2021).Rare estimates of the overstepping degree for the garnet in reaction along Barrovian gradient yield values of about 50°C and 0.2-0.5 GPa (Spear et al., 2014), hence a depth shift of about 10 ± 5 km depending on the thermal gradient.Depending on possible cascade effects, the reaction interval could also be smeared or sharpened compared to equilibrium-based modeling (Pattison et al., 2011).Thus, if the velocity jump in the tomographic model is due to metamorphic reactions, taking into account the reaction kinetics in the velocity calculation would shift the predicted velocity step to greater depths.In this prospect, calculations from most samples (the Corsica kinzigite, the Argentera amphibolite, the Sesia gneiss, the Canavese blue-schist and the Taou Blanc chlorito-schist) yield velocity jump with amplitude similar to the observed profile but at shallower depths.On the opposite, a velocity change related to eclogitization would occur at greater depths (from 50 to 80 km).Kinetics effects could therefore explain why the amphibolites-to-granulites are imaged deeper than expected, not how the eclogite-in could occur at shallower levels.

Implications for Other Seismic Imaging Results
Among the available V p models using local earthquake tomography, the one along the CIFALPS transect (Solarino et al., 2018) shows intriguing properties in the European LC and the dipping slab.At 30-40 km depth, from the Penninic Front (PF) to the Ivrea body, the increase in V p from 6.70 to 7.60 km/s (Solarino et al., 2018), associated with an increase in V s from 3.90 km/s to 4.20 km/s (Lyu et al., 2017), is interpreted as a progressive change in the metamorphic assemblage during subduction, associated with a change in density from 2.90 to 3.33 kg/m 3 (Solarino et al., 2018).
The present data set is in agreement with the results of the P-wave velocity model, suggesting the absence of a major inherited lithological boundary and an amphibolites-facies heritage for the European LC.V p calculations for lithologies considered relevant for the amphibolites-to-granulites transition scenario (i.e., the Canavese mylonitic blueschist, the Val Sesia gneiss, the Argentera amphibolite) yield values between 6.8 and 7.20 km/s at 45 km depth.They are too low to explain the V p signature at 45 km in the Alpine slab in V p tomography images (Paul et al., 2022;Solarino et al., 2018) but could be relevant for the low velocity dipping body imaged at the top of the dipping panel ("c" domain in Solarino et al., 2018) located at the same horizontal distance to the PF Thrust as the low velocity bulge of the slab in V s tomography.A further discussion of these apparent discrepancies is actually not possible without a consistent definition of the slab limits, which significantly vary from a model to another.A consistent inversion of V p , V s and their ratio, together with a depth migration of receiver functions in the same velocity model, would actually feed a further discussion of the petrological significance of velocity contrasts along the Alpine slab top.

Conclusion
In this paper, we calculated the seismic properties of eight crustal rocks, relevant as analogs of the lower crustal part of the Alpine slab, and compared them with the seismic profile of the Alpine slab in a recent V s ambient noise tomography model (Nouibat et al., 2022).
The well-defined increase in velocity at 25-45 km depth in the upper part of the Alpine slab in the tomographic model can be explained by the entrance of a felsic to intermediate composition LC panel in the HP granulites facies, leading to the consumption of low nominal seismic velocity minerals, such as amphibole and/ or biotite, and the growth of mineral phases with higher seismic velocities, such as garnet and pyroxene.The best felsic and intermediate analogs in the field are a kinzigite from Monte San Petrone, Corsica, a gneiss from the IL and a blueschist mylonite from the Lanzo Canavese contact.Direct measurements of seismic wave velocity and calculations of seismic properties from CPO measurements will help clarify the effect of anisotropy in our samples.No inherited lateral lithological contrast is needed to explain the velocity profile from the tomographic model.The lower level of the Variscan crust had to be still significantly hydrated to react when entering the HP granulite-facies field.This modeling implies a partially re-equilibrated thermal gradient in the Western Alps, reaching 800°C at 45 km depth, which is consistent with available evidence on the Alpine thermal state at depth.At these temperature, some chlorite-bearing ultramafics rather than serpentinites could explain the seismic signature of the slab top edge.Unless the Alpine slab hydration state is more hydrated than our field analogs, the present study implies a significant reaction overstepping for the amphibolites-to HP granulites-facies transition to be imaged at 45 km. and the "RGF Alpes" project funded by the BRGM.We also thank JB Jacob for sampling in the Belledonne massif and Laurent Jolivet for the field in Corsica.We are grateful for the stimulating discussions with Claudio Rosenberg and Nicolas Bellahsen.Sascha Zertani and an anonymous reviewer are thanked for their remarks and suggestions which have greatly helped to improve this work as Claudio Faccenna for his editorial handling.

Figure 1 .
Figure 1.Location of representative depth velocity profile on CIFALPS transect (a) CCP receiver-function stack, (b) S-wave velocity tomography (data Nouibat et al., 2022) and (c) the associated uncertainties, for (a) the European plate at the front of the mountain belt and (b) the Alpine slab beneath the mountain range.Abbreviations: DM: Dora Maira, PF: Penninic Front.

Figure 2 .
Figure 2. S-wave velocity profiles as a function of depth, in the European plate (a) and along the Alpine slab (b), extracted from the tomographic CIFALPS transect (Figure 1).The geophysical Moho corresponds to the velocity contour 4.3 km/s, selected by Nouibat et al. (2022) as the best Moho proxy for European crust out of the Alpine wedge.
Mineral Phase Maps to S-wave Velocity Maps

Figure 5 .
Figure5.S-wave velocities maps of eight rocks and their evolution along the subduction (8°C/km) and collision (30°C/km) gradient, compared to the V s range of the Alpine slab and the one of the European plates.Samples are divided into three groups according to the topology of their seismic properties.(a) group 1: governed by the plagioclase-out reaction; (b) Group 2: one velocity jump due to the amphibolites-to granulites-facies transition; (c) Group 3: two velocity steps due to the amphibolitesgranulites transition and to the eclogite-in reactions.S-wave velocity maps are calculated from the aggregate mode (derived from thermodynamic models, and from thermoelastic parameters) using theHacker and Abers (2004) seismic property model.Only the phases representing more than 1% of the rock are considered in the calculation.

Figure 6 .
Figure 6.Saturation curves of selected samples and evolution of S-wave velocities of the Corsica kinzigite belonging to group 3, along the subduction (8°C/km) and collision (30°C/km) gradient, for three states of hydration, compared to the V s range of the Alpine slab.

Figure 7 .
Figure 7. End-member thermal regime sketches from the continental subduction of the Alps to present-day conditions and the PT profile associated.Cross sections are at the same vertical and horizontal scales.Abbreviations: A: amphibolite facies; Br: Briançonnais; D: Dauphinois; E: Eclogites facies; G. HP : High Pressure Granulites facies; LC: lower crust; LPT: Ligurian Piemontais Thrust; PF: Penninic Front; RMF: Rivoli-Marene deep Fault; UC: upper crust.Reaction curves in green refer to the ultramafic system.