The Theodul Glacier Unit, a slab of pre-Alpine rocks in the Alpine meta-ophiolite of Zermatt-Saas, Western Alps

The Theodul-Glacier-Unit (TGU) is a 100 m thick and 2 km long slab of pre-Alpine schist, gneiss and mafic rocks tectonically emplaced in the eclogite-facies Zermatt-Saas meta-ophiolite nappe (ZSU). The meta-sedimentary rocks occur mostly as garnet-phengite schists with locally cm-sized garnet porphyroblasts. The metavolcanic basic rocks are present as variably retrogressed eclogites showing a continental basalt signature and contain abundant zircon, which is unusual for basalts. The zircons dated with the U–Pb system yield an upper intercept age of 295 ± 16 Ma and a lower intercept age of 145 ± 34 Ma. The early Permian age is interpreted to represent the age of high-grade granulite facies metamorphism, evidence of which is also preserved in the cores of garnet porphyroblasts of the Grt-Ph schists. The lower intercept age corresponds to the time of continental breakup and the initiation of the Tethys in the Mid-Jurassic; these events may have created the TGU as an extensional allochton. Eclogite facies metamorphism recorded by the TGU rocks occurred during Alpine subduction at 57 Ma, the Lu–Hf age of TGU eclogite garnets. The TGU reached a depth of about 53 km at P–T conditions of 1.7 GPa and 520 °C derived from both, eclogite and Grt-Ph schist. This is in contrast to the ZSU surrounding the TGU with a reported subduction depth of more than 80 km at 43 Ma. It is proposed here that TGU and ZSU were subducted separately out of sequence. After juxtaposition of the two units during late Alpine thrusting and folding forming the present day geometry of nappes in the Zermatt-Saas region both units were progressively metamorphosed to about 650 MPa and 470 °C. This late prograde metamorphism at 34 Ma produced oligoclase + magnesio-hornblende in the matrix of Grt-Ph schists and eclogites. The derived TGU data document a complete Wilson Cycle.


Introduction
The very rapidly retreating Upper Theodul Glacier south of Zermatt close to the Swiss-Italian border exposed a slab of strikingly rusty-weathering rocks in the metamorphic Mesozoic ophiolite of the Zermatt-Saas Unit (ZSU) close to the major thrust contact to the cover unit, the Combin Unit, above ( Figs. 1 and 2). This about 2 × 0.1 km large slab is made up of rocks that are not typical of the ZSU (Weber and Bucher 2015). The slab represents an exotic tectonically emplaced fragment of continental rocks and has been labeled Theodul Glacier Unit (TGU) (Weber and Bucher 2015). The rock assemblage of the TGU is unique to the Zermatt region and the origin and significance of the TGU is scarcely known. This paper presents new data from eclogite and garnetphengite schist of the TGU, including zircon ages from the eclogite. The data show that the TGU rocks have been metamorphosed under granulite facies conditions in the Permian. The continental breakup and the initiation of the Tethys in the Mid-Jurassic (Dogger) are reflected by the composition of the zircons. This event created the TGU as an extensional allochton. Alpine subduction  Bearth 1967). The Mischabel and Monte Rosa nappe are Penninic crystalline basement (European plate), the Dent Blanche nappe represents Austroalpine continental basement (Apulian plate). TGU Theodul Glacier Unit (Weber and Bucher 2015) see detailed geological map Fig. 3, LdC Lago di Cignana coesite locality (Reinecke 1991(Reinecke , 1998, eclogitic pillow lavas at Pfulwe locality (Bucher et al. 2005), eclogite facies meta-gabbro at Allalinhorn (Bucher and Grapes 2009) tectonic elements (Figs. 1 and 2). The nappe consists of three major units (Argand 1908;Bucher et al. 2003): (a) the Arolla unit consists of banded granitoid gneisses and massive granites, (b) the Valpelline unit is made of prealpine high-grade metamorphic rocks including garnetcordierite-sillimanite gneiss, tremolite-diopside marble and garnet amphibolite and (c) large masses of Permian gabbro (Dal Piaz et al. 1977;Manzotti et al. 2017Manzotti et al. , 2018. The Dent Blanche nappe system shows complex internal Alpine deformation (Bucher et al. 2003(Bucher et al. , 2004 and can be subdivided into various sub-nappes (Dal Piaz et al. 2001;Manzotti et al. 2014;Dal Piaz et al. 2015).
Other regional and local units: The Combin Unit (CU) consists predominantly of Mesozoic meta-sediments and meta-ophiolites, mainly calcareous micaschists and greenschists, but also Triassic marble and quarzite (Bearth and Schwander 1981) (Figs. 1 and 2). It preserves rare blueschist facies assemblages but no eclogites (Bucher et al. 2004). The CU basal thrust ( Figs. 1 and 2) is, therefore, a major discontinuity in metamorphic grade and juxtaposes two units with contrasting subductionexhumation histories during late collision and nappe stacking. The Tsaté nappe with relics of metamorphic ophiolitic material is located between the CU and the DBN (Escher et al. 1993) labeled as upper ophiolites on Fig. 2. The metabasic rocks of the Tsaté nappe do not show eclogite facies assemblages.
Slices of continental rocks comprising also pre-Tethys metasediments occur between ZSU and CU, for instance the Etirol Levaz slice (Ballèvre et al. 1986;Fassmer et al. 2016) south of the Lago di Cignana locality (LdC on Fig. 1).

Structure
The Theodul Glacier Unit (Fig. 1) Fig. 2 Overview over the field area as seen from point 3030 to the NW. Matterhorn belongs to the Dent Blanche nappe from the overriding plate ( Fig. 1). Below the main Austroalpine thrust follow the upper ophiolites (Tsaté nappe: Escher et al. 1993) and various units hereafter collectively referred to as Combin Unit (Dal Piaz 1965;Bearth 1967). The top of the Zermatt-Saas ophiolite is at the major thrust fault at the foot of Hirli. The rusty-brown weathering garnet-phengite schists and eclogites in the foreground form the Theodul Glacier Unit (TGU) of this study, named after the Upper Theodul Glacier to the south (left) of the outcrops. The topographic points are also shown on the geological map ( Fig. 3) Zermatt-Saas ophiolites about 100 m below the basal thrust of the Combin Unit (Fig. 2). The TGU slab is imbricated and has been transected by secondary thrusts and ZSU material, predominantly serpentinites but possibly also eclogite ( Figs. 3 and 4). The basal main thrust fault separating the TGU from ZSU cuts minor thrusts in the ZSU. The slab is covered by ZSU serpentinites along a top thrust (Fig. 3). The serpentinite outcrops near point 2924 (green stars on Fig. 3) expose two types of meta-rodingites (Li et al. 2004b(Li et al. , 2008. A separate slice of garnet-biotite gneiss rests on the TGU and ZSU rocks (Fig. 3).

Rock assemblage
The three rock types of the TGU are tectonically interleaved and the variability at outcrop scale is much larger than shown on the simplified geological map (Fig. 3). Garnet-phengite schist represents the most abundant rock type of the TGU (Fig. 4). The cm-large garnet porphyroblasts show a complex internal chemical structure. The core regions of the garnets formed under granulite facies conditions whilst a younger generation of garnet material in the schists can be related to eclogite facies metamorphism (Bucher et al. 2019). Lenses, boudins and blocks of eclogite from some centimeters to many meters in size are common within the Grt-Ph schists (Fig. 5a) (all abbreviations for mineral names after Whitney and Evans 2010). Banded eclogite in Grt-Ph schist can be broken and rotated (Fig. 5b). The composition of the TGU eclogites is distinctly different from the ZSU eclogites and suggests a within-plate origin of the original mafic rocks rather than being ophiolitic (Weber and Bucher 2015). The eclogites contain abundant Qz and Ph in addition to Grt and Omp in contrast to Qz-absent ZSU eclogites with abundant Pg, Mg-Cld and Gln. The TGU eclogites contain as well a notable amount of Zrn contrary  Klemm 2005). Note that the map is a revised and improved version of the local geological map in Weber and Bucher (2015). Easting and northings according to the Swiss coordinate network to the ZSU eclogites. The eclogite facies metamorphism occurred about 57 Ma ago, in the late Paleocene, as indicated by Lu-Hf garnet-whole rock analyses of TGU eclogites . The Bt gneiss and Bt schist shown on the geological map ( Fig. 3) are not described in this paper.

Methods
The composition of minerals was determined using a CAMECA SX-100 electron microprobe at the University of Freiburg. All quantitative analyses were made using wavelength-dispersive spectrometers. Operating conditions were 15 kV acceleration voltage and 15 to 20 nA beam current with counting times of 10 s and a focused electron beam. Mica was measured with a defocussed beam (5 µm) to minimize the effect of Na and K loss during analysis. The instrument was calibrated for each element analyzed using well-characterized natural materials as standards. Data reduction was performed utilizing PAP software provided by CAMECA.
Whole rock analysis was performed by standard X-ray fluorescence (XRF) techniques at the University of Freiburg, Germany, using a Philips PW 2404 spectrometer. Pressed powder and Li-borate fused glass discs were prepared to measure contents of trace and major elements, respectively. Raw data were processed with the standard XR-55 software of Philips. Relative standard deviations 2σ are < 1 and< 4% for major and trace elements, respectively. Loss on ignition was determined by heating at 1100 °C for 2 h. Rare earth elements (REE) analyses were performed by sodium peroxide fusion ICP AES at SGS Minerals Services in Lakefield, Ontario Canada.
P-T conditions during the different stages of metamorphism have been deduced from thermodynamic models using Theriak/Domino software of de Capitani and Petrakakis (2010), and thermodynamic data from Berman (1988) (JUN92 data and updates). The thermodynamics of Grt, Cpx and white mica solutions follow the models of Berman (1988), Meyre et al. (1997) and Keller et al. (2005), respectively. a b d c point 3002 Fig. 4 a NE-face of point 3030. Low ground: Light-gray foliated meta-gabbro of the ZSU with well-preserved omphacite. Top part of wall: Rusty-weathering Grt-Ph schists (sampling site of KB860 and OK-Z133) and Bt-gneiss (Fig. 3). The basal thrust of the Theodul Glacier Unit in the middle of the wall and is well marked with small serpentinite lenses (e.g. at the position of the two geologists in white circles). b Point 3002 seen from north (Figs. 2 and 3). Basal thrust of TGU marked with red dashed line (ZSU above the lake, TGU at the top). Along the thrust several small serpentinite lenses occur. Rusty TGU rocks are predominantly Grt-mica-schists. Eclogites and Bt-gneiss are greyish colored. c Garnet-phengite schist at outcrop scale. Note cm sized and euhedral undeformed garnet (hammer head is 12 cm). d Close-up of framed area on c Dating was carried out with U-Pb isotope dilution thermal ionization mass spectrometry (ID-TIMS). Zircon grains were selected under a binocular microscope and subjected to chemical abrasion (Mattinson 2005) before spiking with a 202 Pb-205 Pb-235 U tracer, dissolution, and mass spectrometry, following the procedure of Krogh (1973) with modifications described in Corfu (2004). The data were corrected for blanks of 0.1 pg U and ≤ 2 pg. The remaining initial Pb was corrected using compositions calculated with the model of Stacey and Kramers (1975). Plotting and regressions were done with the Isoplot software package (Ludwig 2009). The decay constants are those of Jaffey et al. (1971). Uncertainties in the isotope ratios and the ages are given and plotted at 2σ.

The eclogites
Samples of TGU eclogite have been collected and studied across the TGU (Fig. 3) during fieldwork from 2005 to 2015 (Weber and Bucher 2015). Sample SW9 used for Zrn dating comes from the locality marked with a blue star on the geological map (Fig. 3). The exact location has the coordinates: 620,650 easting, 91,470 northing (Swiss coordinate network).
Sample SW9 contains the minerals Grt, Cam, Pl, Chl, Ph, Pg, Rt, Ttn, Qz, and Zrn. The early Alpine highpressure assemblage Grt + Omp has been replaced by the later assemblages with Pl, Cam, Chl and Ttn. The rock contains also subordinate amounts of Qz, Pg and Zrn. Small grains of Ap, Ep and sulfide (mostly Py) are also present. There are no Omp relics preserved in this sample, in contrast to many other partially retrogressed basic meta-volcanic rocks from the TGU (Weber and Bucher 2015). The texture of the SW9 rock is dominated by coarse well-aligned optically zoned Cam. Pl forms an equilibrated texture with Cam. Grt is euhedral and strongly poikilitic. The Grt are typically rimmed with a thin seam of late Chl (see also Fig. 5c). The mineral inclusions comprise Rt, Qz, and Pg and show sigmoidal trails indicative of rotation during growth. Grt has no Cam inclusions.
Bulk K 2 O is relatively low in SW9, which is reflected by low modal Ph. It occurs together with abundant Pg as a fine-grained mineral in the Cam-rich matrix. Abundant Chl occurs in association with Ph and Pg and occasionally as a late replacement of Grt. In the foliated Cam-mica matrix of the rock, Chl locally grows across the foliation.

Composition of the SW9 eclogite and its minerals
The chemical composition of the mafic metavolcanic sample SW9 is given for major and some trace elements on Table 1. The composition of SW9 is similar to those of other mafic rocks including all eclogites of the TGU. The TGU metavolcanic rocks derive from typical "within plate basalts" in contrast to mafic rocks of the ZSU, which are "MOR or OI basalts" (Weber and Bucher 2015).
Garnet is compositionally zoned with Grs-rich Prppoor cores and a mantle (rim) richer in Prp and lower in Grs (Table 2). There is more Sps in the core than Prp. Sps decreases rapidly towards the mantle where it is very low. This features are typical of prograde zonation in Grt. The core area is about 40% of the Grt diameter. Consequently, the core stands for 6 vol.% of the total Grt volume, the  (Fig. 2). a eclogite lens within Grt-Ph schist. b Rotated blocks of banded and foliated eclogite in Grt-Ph schist matrix. c eclogite with ~ 10% modal euhedral garnet of 2-5 mm in diameter in a matrix of relic Omp, Cam, Ph, Pg, Chl, Ep, Qz, and accessories mantle and rim represent 94 vol.% of the Grt. There is about 10 vol.% modal Grt in SW9 (Fig. 5c). Grt core material represents 0.6 vol.% of modal Grt. Therefore, it is not necessary to correct the bulk rock composition for the fractionation into the Grt core for the P-T model for the Grt mantle and rim growth presented below.
The Si content of phengite reaches 3.35 atoms per formula unit (a.p.f.u.). Phengite contains about 10 mol% Pg component and X Mg is 0.72 in Cel-rich Ph and 0.63 in late Ms-rich Ph (Table 3). Chlorite occurs in different generations and compositions, the first being Mg-rich, the last Fe-rich (Table 3). All amphiboles are Ca-amphiboles but differ considerably in composition (Table 4) and can be classified as actinolite and magnesio-hornblende (Hawthorne et al. 2012). The Al-content varies from 0.5 in Act and increases to 1.5 a.p.f.u. in Mhb. Between 0.7 and 2 wt% Na indicates that Cam developed from Omp originally present. Plagioclase is an oligoclase with An ~ 21 (Table 5).

P-T model for eclogite SW9
P-T estimates have been obtained utilizing the Theriak/ Domino software (de Capitani and Petrakakis 2010) using rock and mineral composition data presented above (Fig. 6). The rock composition (Table 1) has been simplified to the model system as follows: TiO 2 is predominantly present in rutile. A small amount of TiO 2 is present in titanite formed during retrogression and has been ignored. 0.2 wt% MnO has been ignored. The rock contains 0.46 wt% P 2 O 5 present in Ap (Ca 5 (PO 4 ) 3 (OH)). Thus a stoichiometric amount of 0.43 wt% Ca has been subtracted from the rock composition given in Table 1. The remaining components are included in the phase equilibrium model.
The first Grt appears at about 500 °C and 1.5 GPa in rocks with the composition of SW9 (Fig. 6a) along a subduction geotherm of 9.5 °C/km. The computed first Grt has the composition Grs 35, Prp 8, Alm 57 (Fig. 6b). The core of the measured SW9 Grt has Grs 34 but a much lower Prp of 4. This mismatch is related to the presence of 6 mol% Sps in the core, which is not adequately modeled here (Konrad-Schmolke et al. 2008). The onset of Grt growth strongly fractionates MnO into the core of the new Grt thus expanding its stability field to lower temperature. This results in a lower temperature for the first Grt with a lower Prp of the core.
The mantle and rim of Grt representing > 90 vol.% of the Grt contains 30 mol% Grs component, 8 mol% Prp, 62 mol% Alm and only traces of Sps (Table 2). For this Grt composition the three isopleths intersect at 503 °C and a slightly higher pressure of 1.63 GPa (point 1 Fig. 6b). However, the Grt isopleths intersect at a second point (point 2 Fig. 6b) at 510 °C and 1.3 GPa. Yet the Grs isopleth are narrow spaced and vertical at point 2, which would lead to strongly Grs zoned Grt during prograde metamorphism. The SW9 garnet shows a very small Grs variation from 34 in the very core to 30.4 mol% in the outer rim zone (Table 2) in agreement with the orientation and spacing of the Grs isopleths  at point 1 (Fig. 6b). Modeled Si in Ph is close to 3.38; measured Ph has a Si of 3.35 (Table 3). The predicted stable assemblage at these P-T conditions is: Grt-Omp-Ph-Pg-Chl-Qz-Lws. The assemblage is present in SW9 except for Omp and Lws, which have been replaced mostly by Cam during greenschist facies recrystallization. At point 1 Gln is not part of the stable assemblage but it is present at lower temperature before the first Grt forms. Plagioclase (oligoclase) in the rock matrix is not predicted to be stable under these P-T conditions. The computed modal Grt of 7 vol.% matches modal Grt of ~ 10 vol.% in the sample SW9 reasonably close. The observed composition of the Grt suggests that the rock has not been subducted to conditions beyond point 1 (Fig. 6a) corresponding to 53 km depth (rock density 3000 kg/m 3 ). With increasing P-T Grs decreases and Prp increases significantly and reach 24% Grs and 23% Prp at 600 °C and 2 GPa, for example. At these conditions predicted modal Grt would be > 40 vol.% in contrast to the observed ~ 10 vol.%. The matrix assemblage Pl An 20 + Mhb can be related to a later post-eclogite recrystallization. The modeled P-T conditions during this matrix recrystallization are ~ 450 °C and 550 MPa (Theriak model). Computed Pl has An 25 and Ca-Amp contains 81 mol% Prg in agreement with analyzed Pl and Amp (Tables 4 and 5). Computed X Mg in Chl of 0.5 compares with the measured X Mg = 0.4 for Chl in the matrix of SW9 (Table 3).

Characteristics of SW9 zircons
The population is composed largely of equant to shortprismatic grains, generally highly subrounded. Cathodoluminescence images reveal complex textures, in part with straight, regular zoning ( Fig. 7a) or with more random mosaic-type textures (Fig. 7b). The grain in Fig. 7c is characterized by a CL-dark internal domain surrounded by a CL-bright outer zone showing a very faint regular growth zoning. All zircons show variable evidence of reworking resulting in dissolution, recrystallization and the formation of veins, locally thin CL bright overgrowths and zones of disrupted and convolute zoning. This feature is most strongly developed in grain of Fig. 7d, both in the CL-dark core and in the CL-bright rim.

U-Pb results
The data include one fraction of 8 small round grains (too small for single-grain analysis), and 7 single grain analyses ( Table 6). The data show a high degree of dispersion with U-Pb ages ranging from about 290 to 150 Ma (Fig. 8). Six of the analyses define a line (with an MSWD of 3.1) having intercept ages at 295 ± 16 Ma and 145 ± 34 Ma. The oldest analysis deviates to the right of the line, suggesting the presence of an older component probably related to the magmatic stage. However, the extrusion age of the basalts could not be quantified and remains unknown. The granulite facies event at 295 Ma whiped out the magmatic age. The two youngest analyses also deviate from the line. They represent grains that became highly turbid after chemical abrasion, the youngest one an isolated tip. They were analyzed to test the lower intercept age of 145 Ma, whether it could be confirmed and whether it might correspond to an event that formed new zircon. The results do not provide a conclusive answer to this question, but they show that at least some of the discordance is due to younger overprints, likely during the Alpine events.
Considering the combination of observed textures and the complex, but broadly coherent data, it can be concluded that the upper intercept age of 295 ± 16 Ma corresponds to the time of high-grade metamorphism, consistent with the mosaic textures seen in part of the zircon grains (Fig. 7b). The analysis with the slightly older age may reflect an original magmatic zircon component, possibly represented by regular zoned grains such as seen in Fig. 7a. The reasons for the subsequent strong disturbance of the U-Pb systems are less clear, but it was likely related to the extensive reworking of zircon documented in textural features such as seen in Fig. 7b, c. The lower intercept age of 145 ± 34 Ma suggests that the event could have been associated with hyperextension processes during continental breakup and the initiation of the Tethys in the Mid-Jurassic (Dogger). The zircon U-Pb systems were in part also affected during the Alpine events that formed, and then retrogressed, the eclogite.

The garnet-phengite schists
Garnet data from the TGU garnet-phengite schist (Figs. 4c, d) have been presented by Bucher et al. (2019). Two porphyroblastic Grt samples, KB860 and OK-Z133, exhibit particularly well-developed internal chemical structures in garnet and have been studied in detail. Here, we supplement additional data from TGU Grt-Ph schists and refine the P-T history deduced by Bucher et al. (2019). There are new P-T data for the Grt-Ph schists for the post-eclogite stages and matrix-recrystallization.    Furthermore, there are previously unpublished rock and mineral composition data presented and discussed in this section (e.g. REE and traces of the rock). Since the study by Bucher et al. (2019) focussed on Grt, there are few other mineral analyses given in that paper. Here, the composition of matrix minerals are presented in Tables 9, 10, 11, 12 and used for an improved derivation of the post-eclogite metamorphism. The Grt-Ph schists contain euhedral Grt porphyroblasts (Fig. 4c, d) occurring in a matrix of Ph, Pl, Chl, Pg, and modally subordinate Ep, Qz, Ttn, Rt, rare Amp, and Fe-sulphide (Fig. 9a). The Ph-and Chl-rich rocks are well foliated (Ph, Pg, Chl > 70% modal). The texture of some samples bears evidence for two generations of Ph (Fig. 9b).
Large porphyroblastic Grt crystals consist of an aggregate of angular and blocky fragments of first generation Grt with a second generation of Grt material between the fragments and along the rims of the porphyroblasts. The first generation of Grt is rich in Prp component and formed by pre-Alpine granulite facies metamorphism. The late generation of garnet is low in Prp and rich in Grs component. It was formed during Alpine eclogite facies metamorphism (Bucher et al. 2019).
The dominant mica is a fine-grained Ph forming the schistose structure. It occurs intergrown with Pg and also occurs as fine inclusions within Pg. Abundant Chl is present in association with Ph and Pg and locally grows across the foliation. Epidote appears as small (< 50 µm) euhedral grains with trapezoidal cross section locally overgrowing corroded REE Ep, allanite (Fig. 9b). Plagioclase is a member of the matrix assemblage together with the white micas and Chl. Amphibole occurs locally as a minor phase associated with Pl. The presence of Chl-Ab pseudomorphs after Amp rimmed by Act suggest that Gln was present in the rock at some earlier stage (Fig. 9c). Accessory Ttn locally replaces Rt.

Rock and mineral composition 6.1.1 Rock
The major element composition of the Grt-Ph schist sample KB860 is given in Bucher et al. (2019). KB860 is a terrigenous clastic rock and classifies as shale. The relatively high CaO suggests that the shale contained carbonate (calcite) prior to the first metamorphism and that the sedimentary protolith classifies as calcareous mudstone.
Here we present new REE and other trace element concentration data analyzed in addition for a complete characterization of the modeled material (Table 7). The total rare earth element (REE) content of 293 ppm is enriched relative to chondrite values (Fig. 10a). The chondrite-normalized REE pattern decreases regularly from La to Sm but it becomes more flat at the heavy rare earth elements (HREE) and there is a significant Eu depletion. High field strength elements are relatively abundant including Ti (TiO 2 = 1.16 wt%), Y (45.4 ppm), Zr (197 ppm), Nb (26 ppm), and Th (21.3 ppm) (Fig. 10b). The REE and trace element diagrams show patterns similar to, but higher than those of the North Atlantic Shale Composite (NASC) indicating that the Grt-Ph schists derive from a typical shale. The TGU metasediments are unusually rich in REE in contrast to both groups of the ZSU metasediments presented and distinguished by Mahlen et al. (2005) (Fig. 10a). REE patterns from metasediments of the Valpelline series of the Dent Blanche nappe (Mahlen et al. 2005), a potential source material of the TGU Grt-Ph schists, are parallel to the TGU pattern but also contain less REE than the TGU rocks (Fig. 10a). The REE data suggest another, hitherto unknown source for the calcareous shales of the TGU.  Gromet et al. (1984); CC (continental crust) from Rudnick and Gao (2003), ZSU metasediments (Group 1 and 2 from Mahlen et al. 2005), paragneiss from the Valpelline series of the Dent Blanche nappe (Mahlen et al. 2005). b) Multiple trace element diagram (spider diagram). Primitive mantle data for normalization from McDonough and Sun (1995). For comparison NASC (North American shale composite) from Condie (1993)

Garnet
A detailed chemical characterization of the garnets in TGU Grt-Ph schists has been presented in Bucher et al. (2019). Average compositions of granulite facies Grt cores and eclogite facies Grt rims and veins from 87 analyses are given in Table 8. The highest Prp content of typically 28 mol% Prp is found in the center of fragments of granulite facies Grt. Grossular concentration is 5.5 to 6.0 mol%. Eclogite facies Grt between the Prp-rich fragments is rich in Grs containing up to 35.8 mol% Grs component. Pyrope with about 8.5 mol% Prp, is distinctly lower than in granulite facies fragmented cores (28%). The contact between the two Grt is very sharp and defined by the Grs content, which jumps from 6 to 35 mol% over a distance of only 2-3 µm (Bucher et al. 2019).

Other silicate minerals
The composition of Ph varies considerably between and within samples (Table 9). Highest measured Si is 3.4 a.p.f.u. in a Ph with low X Fe but typical values are Si = 3.2 a.p.f.u. Pure Ms occurs as inclusions in Pg and as late Ms growing across the foliation. Paragonite of constant composition contains about 6 mol% Ms component. All Chl is Clc with X Fe of about 0.45 for matrix Chl (Table 10). Epidote shows strong chemical zoning ( Fig. 9b; Table 11). On BSE images a visible bright core indicates the presence of corroded allanite, the REE-rich variety of Ep. Some cores consist of Fe-free REE-Zo (two single point analyses on Table 11). The darker euhedral rims (BSE images, Fig. 9b) are composed of stoichiometric Ep (Clz). The two types of Amp present can be classified as Hbl and Act (Hawthorne et al. 2012), with Na of 0.75 and 0.16 a.p.f.u. respectively (Table 12). The Hbl can be characterized as ferrian subcalcic Mhb. The rim of Chl + Ab pseudomorphs after Gln (Fig. 9c) consists of Act overgrown by Hbl. Most Pl occurring in the matrix of samples KB860, OK-Z131 and OK-Z133 is oligoclase with An 13 -An 18 . In all samples some Ab is present particularly in pseudomorphs after Gln.

P-T model for metamorphism of the Grt-Ph schists
The following section summarizes and discusses the P-T estimates for the Grt-forming stages from Bucher et al. (2019) and presents updated P-T condition for the posteclogite matrix assemblages.

Granulite facies
The cores of the porphyroblastic Grt formed at granulite facies conditions near 780 °C ± 30 and 670 MPa ± 50 (Bucher et al. 2019). The predicted assemblage at these conditions and in the presence of a graphite-controlled fluid is Grt-Kfs-Pl-Bt-Crd-Sil-Qz. Garnet with graphite inclusions (+ Qz) is the only mineral from this assemblage that survived later hydration and recrystallization.

Eclogite facies
The Grs-rich second Grt generation formed at conditions of the eclogite facies, 530 °C and 1.7 GPa. The model assemblage at this P-T is Grt-Omp-Ph-Pg-Qz-Lws. The modeled assemblage differs from the observed assemblage in that neither Omp nor Lws, are present in the rock. It is possible that Omp + Lws were originally present during eclogite facies growth along a prograde subduction path, but were later removed from the matrix during complete greenschist facies recrystallization of the rock matrix.
The model fails to predict the presence of glaucophane at the eclogite stage or pre-eclogite stage for which there is textural evidence in the form of Ab-Chl pseudomorphs (Fig. 9c) produced by the reaction 1: (1) 1 Gln + 1 Pg + 1 Qz + 2 H 2 O = 1 Chl + 3 Ab

First greenschist facies event
The assemblage Chl-Ab-Pg-Ms-Act-Qz is characteristic of greenschist facies conditions. The textures suggest that the greenschist facies overprint followed the eclogite facies. The P-T conditions of this stage probably where close to 350 °C and 300 MPa (Bucher et al. 2019).

Second greenschist facies recrystallization
In the recrystallized matrix of the schists, Pl with An 18 and Cam with Prg 70 record prograde metamorphism up to the greenschist to amphibolite facies boundary. The Ep/Clz of the matrix belongs to the same assemblage together with both white micas and Qz. A P-T estimate of 470 °C and 650 MPa for this last and prograde metamorphic stage results from new Theriak models using the matrix composition given in Bucher et al. (2019). The model reproduces the observed Pl and Cam composition and the matrix assemblage. This metamorphic stage has not been modeled previously and the derived P-T estimate is comparable to the P-T estimate of 550 MPa and 450 °C derived for the SW9 metabasalt given above. The metamorphic stage most likeley reflects the effect of the late Alpine collision phase.

Variscan granulite facies metamorphism
The composition of the cores of porphyroblastic Grt is evidence for granulite facies metamorphism of the metasediments of the TGU. The morphology of Zrn in mafic metavolcanic rocks of the TGU is consistent with a metamorphic granulite facies origin. The upper intercept age at 295 ± 16 Ma (Fig. 8) suggests that the high-grade metamorphism occurred close to the Carboniferous-Permian boundary in the final stages of the Variscan orogeny. High-grade metamorphic metapelites have been reported from the Valpelline Series within the Dent Blanche nappe, the tectonically highest unit of the Zermatt region (Figs. 1 and 2) (Nicot 1977;Gardien et al. 1994;Bucher et al. 2004;Manzotti and Zucali 2013;Manzotti et al. 2014). These rocks carry the assemblage Grt-Pl-Kfs-Qz-Bt-Sil, locally also Crd (Diehl et al. 1952;Bucher et al. 2004;Zucali et al. 2011). The assemblage is identical to the predicted stable assemblage for the Grt core generation in the TGU Grt-Ph schists. Manzotti and Zucali (2013) reported peak metamorphic conditions of 814 ± 40 °C and 600-800 MPa for samples of the Valpelline Series that were then subsequently dated by Kunz et al. (2018). That geochronologic study yielded a Permian age, 280 ± 20 Ma overlapping with our SW9 zircon age within the given errors. The published P-T conditions for the Dent Blanche granulites are identical with the derived conditions of garnet growth for the TGU sample (780 °C ± 30 and 670 MPa ± 50). The high-grade metamorphism in Dent Blanche granulites was related to regional lithospheric thinning during the late Paleozoic Manzotti et al. 2018).

Origin of the TGU and possible involvement in Jurassic extensional processes
The similarities of the TGU rocks with the high-grade metasediments of the Valpelline unit of the overriding Dent Blanche nappe described above may indicate that the TGU represents material from the Dent Blanche nappe in spite of not perfectly matching REE patterns (Fig. 10a). How the material crossed the basal thrust and the underlying units to end up as an isolated slab in the ZSU remains unknown. An process of subduction erosion has been proposed for the Allalin gabbro within the Zermatt-Saas ophiolites (Bucher and Grapes 2009). Alternatively, the TGU slab could represent a relic of an extensional allochton, similar to slices of continental basement occurring further south in the ZSU (Beltrando et al. 2010). The slab may have been incorporated into the newly forming oceanic lithosphere during the early stages of continental rifting and extension in Mid-Jurassic time. However, the TGU slab has not been intruded by oceanic basic melts, gabbros and basalts of the ZSU. All contacts between the TGU and ZSU are strictly tectonic, in support of a tectonic process related to either the subduction or collision phase of the alpine orogeny. On the other hand, the granulite facies Zrn from eclogite SW9 indicate that they have been modified by an event, probably the continental breakup, at the Jurassic-Cretaceous boundary (lower intercept age of 145 ± 34 Ma). It is difficult to weight the indications for extensional allochton versus subduction erosion unequivocally.

Alpine eclogite facies metamorphism and relations between TGU and ZSU
Eclogite facies metamorphism has been recorded by both the meta-volcanic rocks and the metasediments. The mafic rock SW9 yielded 503 °C and 1.63 GPa whilst the Grt-Ph schist KB860 supplied 530 °C and 1.7 GPa for the eclogite facies conditions related to Alpine subduction. The 1.63 GPa maximum pressure corresponds to 55 km depth and the temperature of 520 °C gives a geothermal gradient of about 9.5 °C km −1 (for a rock density of 3000 kg m −3 ). The peak conditions of the eclogite facies metamorphism has been dated by Lu-Hf garnet-whole rock isochrons of 56.5 ± 2.7 Ma for the TGU eclogite SW9 and 58.2 ± 1.4 Ma for another sample SW14 . Note that Weber and Bucher (2015) reported P-T conditions of 580 °C and 2.2 GPa for the TGU eclogites and interpreted them as peak conditions of subduction metamorphism. However, the coincidence of our new P-T conditions derived for eclogite SW9 and Grt-Ph schist KB860 suggests that the peak conditions are lower than previously reported. It remains a possibility though that the TGU does not represent a coherent unit. The field occurrence of some of the eclogites as boudins, lenses and rotated blocks in the Grt-Ph schists (Fig. 5a, b) suggests that they have been tectonically emplaced in the schists. Thus there may occur two types of eclogites in the TGU: (i) one with a maximum P of 1.63 GPa identical to the P of eclogite facies Grt formation in the Grt-Ph schists and (ii) one type with maximum P significantly higher (up to 2.2 GPa). The most likely solution resolving the inconsistency is, however, that the eclogite sample providing the 2.2 GPa pressure (LX15 in Weber and Bucher 2015) is a tectonically emplaced fragment from the ZSU (see P-T estimates below).
The maximum pressure recorded by the assemblages of the Zermatt-Saas meta-ophiolites range from 2.2 to 2.4 GPa (Angiboust et al. 2009) for eclogites of the area near Zermatt, from 2.5 to 2.7 GPa (Bucher et al. 2005) near the Pfulwe locality (Fig. 1) and 2.5 GPa for the eclogite facies Allalin gabbro ( Fig. 1) (Bucher and Grapes 2009). The large range of derived pressures may indicate that several separate slices with different subduction -exhumation paths are present within the high-pressure meta-ophiolites. P-T conditions of 2.3 -2.4 GPa and 500 ± 50 °C have been reported for eclogites belonging to the ZSU of the Trockener Steg area enveloping the TGU with its Grt-Ph schists (Weber and Bucher 2015). This maximum pressure of 2.3-2.4 GPa is consistent with the estimates cited above and well within the uncertainty of the applied methods for estimating the pressures. This maximum pressure corresponds to a depth of 80-85 km for the return-point at which the ZSU ophiolites detached from the subducting slab and began the ascent to the shallow crust. At the locality Lago di Cignana (LdC on Fig. 1) about 30 km south of Trockener Steg, Coe has been reported from rocks that may belong to the ZSU (Reinecke 1991). This may indicate that some parts of the ZSU ophiolites have been subducted to more than 90 km depth. A similar return-point has also been suggested for the Pfulwe locality (Fig. 1) on the basis of the stable Cld-Tlc assemblage found in eclogite facies glaucophanites from this locality (Bucher et al. 2005). Estimated maximum temperatures at the return-point range from 520 to 600 °C and by and large correlate with the derived pressures suggesting a coherent subduction geotherm of about 8 ± 1 °C km −1 (Bucher et al. 2005;Angiboust et al. 2009;Bucher and Grapes 2009;Groppo et al. 2009;Rebay et al. 2012).
Lu-Hf garnet whole-rock ages of 56.5 ± 2.7 and 58.2 ± 1.4 Ma for two TGU eclogite samples including SW9 indicate that eclogite facies metamorphism of the TGU occurred in the Palaeocene . The late Paleocene subduction age from the TGU eclogites is considerably older than the Mid-Eocene Rb-Sr phengite ages of 43 -46 Ma from the ZSU meta-ophiolites (Dal Piaz et al. 2001). Sm-Nd and Lu-Hf geochronology on the ZSU indicate protracted garnet growth during prograde metamorphism and showed that the ZSU underwent peak metamorphic HP conditions less than 43 m.y. ago (de Meyer et al. 2014). The ZSU rocks reached greenschist-facies conditions around 38 ± 2 Ma on the basis of Rb-Sr whole rock-phengite isochrons (Amato et al. 1999). A Mid-Eocene age of 44.1 ± 0.7 Ma was also derived from whole Zrn and Zrn rims in eclogite at Lago di Cignana by Rubatto et al. (1998), confirming Sm-Nd isotopic analyses of an essentially unretrogressed eclogitic metabasalt suggest that eclogite facies metamorphism occurred at 52 ± 18 Ma. The large uncertainty is due to the presence of very small amounts of Nd-rich epidote present as inclusions within garnet (Bowtell et al. 1994).
The age difference for the eclogite facies metamorphism between the TGU and ZSU can be related to the northwards progressing subduction process. Subduction ages in the range of 55 to 77 Ma have been reported for eclogite facies rocks of the Sesia Unit south of the Zermatt region and tectonically above the ZSU . Thus deformation and metamorphism migrated from internal to external parts, that is, mostly towards northwest (e.g. Handy et al. 2010). It is not clear to what extent this migration was continuous or episodic (Lister et al. 2001). It implies that the TGU slab was not subducted together with the meta-ophiolites of the ZSU but reached its position within the ZSU (Fig. 3) later probably during late thrusting of the nappe stack.

Final metamorphism during exhumation and nappe emplacement
The first greenschist facies recrystallization replacing e.g. Gln by Chl-Ab pseudomorphs at 350 °C and 300 MPa concludes the subduction related exhumation path. The P-T conditions of 470 °C, 650 MPa for the second matrix recrystallization can been placed on the derived exhumation path for units in the Zermatt area (Weber and Bucher 2015). However, this late prograde matrix recrystallization could also be related to deformation overprinting the high-pressure rocks during Alpine nappe stacking. The derived P-T conditions of the matrix recrystallization correspond the conditions for late Alpine greenschist to lower amphibolite facies overprint recorded by the rocks of the ophiolite nappe Zermatt-Saas (Bearth 1967;Barnicoat et al. 1995;Bucher et al. 2005;Bucher and Grapes 2009;Angiboust et al. 2009). Thus the preferred interpretation of the latest assemblage oligoclase + magnesio-hornblende is that it formed by prograde metamorphism related to the latest deformation of thrusting and backfolding in the late Eocene (e.g. Dent Blanche nappe thrust above Eocene of the Barrhorn Series, Ellenberger 1953).

Conclusions
The geological history of the TGU outlined above is summarized on Fig. 11. Both the metamorphic igneous and sedimentary rocks of the TGU experienced granulite facies metamorphism 295 Ma ago at the Carboniferous-Permian boundary in the context of the last stages of the Variscan orogeny (red circle Fig. 11). The granulite facies zircons from eclogite SW9 indicate that they have been modified by an event at the Jurassic-Cretaceous boundary (not shown on Fig. 11). This may have been the transformation of the TGU to an extensional allochton. In any case the zircon U-Pb systems were further disturbed during the Alpine subduction events (green symbols on Fig. 11).
Lu-Hf garnet-whole rock analyses show that subduction of the TGU occurred in the Paleocene about 57 Ma ago  earlier than the main subduction of the ZS meta-ophiolites at 43 Ma in the Mid-Eocene (Rubatto et al. 1998;de Meyer et al. 2014). The Zrn data presented here have been extracted from Zrn in the same sample SW9 that also supplied the Lu-Hf garnet-whole rock ages making us confident that the Permian, Jurassic and Paleogene ages depict a consistent geological history.
New P-T estimates for both eclogites and metasediments suggest that the TGU reached about 53 km depth during subduction in contrast to 90 km of the ZSU.
During exhumation, the high-pressure assemblages of the TGU rocks were retrogressed to greenschist facies assemblages at conditions as low as 350 °C and 300 MPa (green circle Fig. 11). There is evidence in the Grt-Ph schists of a prograde recrystallization of the matrix at significantly higher P-T of 650 MPa and 470 °C (blue symbols Fig. 11). This latest metamorphic overprint has also been reported from eclogites of the ZSU near Trockener Steg (Fig. 3) with a P-T estimate of 750 MPa and 496 °C for the late event (Angiboust et al. 2009). It is plausible that the large scale thrusting and folding that formed the present day geometry of the nappe stack occurred after the exhumation of the high-pressure rocks. The pressure increase of 450 MPa associated with this late deformation corresponds to a burial of TGU and ZSU by about 15 km, leading to the temperature increase of 130 °C.
The ZSU reached its greatest subduction depth at 44 Ma that is in the Mid-Eocene (Rubatto et al. 1998;de Meyer et al. 2014) and the Dent Blanche nappe (Figs. 1 and 2) was thrusted over Eocene sediments at the Barrhorn (Ellenberger 1953). The Eocene ends at 34 Ma (Gradstein et al. 2004). The proposed age limits are consistent with those given in Steck et al. (2015). This leaves a maximum of 10 Ma for exhuming the high-pressure rocks from 90 km depth to the shallow crust at 300 MPa (10 km) and subsequent large scale deformation including the formation of the nappe stack, the Mittaghorn fold controlling the geometry of the meta-ophiolite nappe ZSU and the Mischabel backfold with which the late overprint at the greenschist to amphibolite facies transition of the TGU is associated with (Fig. 11). Using 50% of this time for exhumation after peak pressure results in a vertical exhumation rate of 16 mm a −1 . This conclusion is consistent with the evidence for unusually high exhumation rates for the ZSU of 10-26 mm a −1 reported by Amato et al. (1999).