Evidence for ancient fractional melting, cryptic refertilization and rapid exhumation of Tethyan mantle (Civrari Ophiolite, NW Italy)

Western Tethyan peridotites exposed in the European Alps show limited amounts of partial melting and mostly fertile compositions. Here we investigate the Civrari Ophiolite (northwestern Italy), which is composed of depleted spinel-harzburgites and serpentinites associated with MOR-type gabbros and basalts. The ultramafic rocks are unique amongst western Tethyan peridotites, showing homogeneous residual compositions after ~ 15% near-fractional melting, lack of pervasive melt percolation and mineral compositions that indicate high-temperature equilibration ≥ 1200 °C. Clinopyroxene chemistry records some of the lowest abundances of Na2O, Ce, and Zr/Hf amongst abyssal peridotites worldwide, suggesting that most abyssal peridotites have been affected by variable degrees of melt retention upon melting or cryptic melt percolation. Locally, cryptic MORB-like melt migration in Civrari peridotites produced orthopyroxene + plagioclase intergrowth around reacted clinopyroxene. These clinopyroxene preserve micron-scale chemical zoning indicating rapid cooling after melt crystallization. 143Nd/144Nd isotopic data indicate that Civrari mantle rocks, gabbros, and basalts are not in isotopic equilibrium. Civrari spinel-peridotites represent a highly radiogenic endmember amongst Western Tethys depleted spinel-peridotites, which together form a partial melting errochron of 273 Ma ± 24 Ma. Ancient near-fractional melting and cryptic melt–rock reaction cause variations in radiogenic εNd and εHf, leading to isotopic heterogeneity of Western Tethys mantle rocks. Such inherited signatures in mantle rocks are most likely to be preserved along (ultra-)slow-spreading systems and ocean–continent transition zones. Electronic supplementary material The online version of this article (10.1007/s00410-019-1603-5) contains supplementary material, which is available to authorized users.


Introduction
The formation of oceanic lithosphere at mid-ocean ridges was spearheaded by the discovery of mid-ocean ridges (e.g., Heezen et al. 1964) and the characterization of ophiolites as remnant oceanic lithosphere thrust upon continental margins (e.g., Anonymous 1972;Decandia and Elter 1972). Over the last 30 years, the diversity of oceanic spreading systems has been underscored by the discovery of (ultra-)slow spreading systems (e.g., Dick et al. 2003) and magma-poor rifted margins (e.g., Boillot et al. 1980). In these cases, oceanic lithosphere is dominated by exhumed mantle, oceanic core complexes and minor basalts and gabbros, with extension being accommodated by the exhumation of mantle to the ocean floor along detachment faults (e.g., Dick et al. 1981;Tucholke et al. 1998). Field studies have pointed out numerous similarities between modern ultra-slow spreading systems, magma-poor rifted-margins and ophiolites in the Western Tethys (e.g., Lagabrielle and Cannat 1990;Lagabrielle et al. 2015;Manatschal and Müntener 2009).
Constraining the petrological and geochemical characteristics of the convective upper mantle has primarily centered on the study of exhumed mantle domains along rifted margins and (ultra-)slow spreading systems (e.g., Elthon 1992;Dick 1989;Snow et al. 1994; Rampone and Hofmann 2012;Picazo et al. 2016;Warren 2016). Although abyssal peridotites were initially thought to be simple residues of adiabatic decompression fractional melting (Johnson et al. 1990), trace element abundances of clinopyroxene and abundance of plagioclase-bearing peridotites indicate that abyssal peridotites record varying degrees of diffuse melt percolation and melt retention (e.g., Elthon 1992;Dick 1989;Brunelli et al. 2006;Warren et al. 2009). Moreover, isotopic studies of mantle rocks exhumed along rifted margins and (ultra-) slow spreading systems have demonstrated the chemical and isotopic heterogeneity of exhumed mantle rocks at small (< meter) and large scale. These data have been primarily explained as a result of the inefficiency of mantle convection, allowing for ancient (> 1 to 2 Ga) refractory mantle domains to persist in the convective upper mantle (e.g., Tribuzio et al. 2004;Cipriani et al. 2004;Stracke et al. 2011;D'errico et al. 2016). Alternatively, it was proposed that depleted mantle domains along ocean-continent transition zones (OCTs) preserved in Western Tethys ophiolites might reflect "ancient" crust-forming events (Rampone et al. 1998;Müntener et al. 2004;McCarthy and Müntener 2015). However, melt percolation in exhumed mantle domains along OCTs and (ultra-) slow spreading systems (e.g., Müntener et al. 2010;Warren and Shimizu 2010) has the tendency to obscure the original composition of depleted mantle domains, making it difficult to relate specific mantle domains to a particular tectonic setting and magmatic processes.
We present a petrological study of one of the most depleted peridotites in the Western Tethys. The Civrari Ophiolite consists of serpentinites and highly depleted residual high-temperature peridotites. The depleted nature of the Civrari mantle clinopyroxene, affected by near-fractional melting, indicates that residual abyssal peridotites do not reflect simple residues after partial melting. We also show how cryptic MORB-type melt percolation occurs at shallow depth, leading to grain-scale chemical zonation of clinopyroxene and occasional plagioclase and apatite saturation. We discuss the possible causes of highly radiogenic Hf-Nd signatures in mantle rocks and suggest that the lack of isotopic equilibrium between mantle rocks and associated magmatism along (ultra)-slow spreading systems and ocean-continent transition zones might be caused by ancient near-fractional melting events and subsequent accretion to the subcontinental mantle.
The Civrari Ophiolite is a sliver of the Piemontese Ophiolites and is located to the West of the Lanzo massif. It consists mainly of serpentinites associated with a thin sedimentary cover of micaschists, calcschists, and quartzites. Field observations show that mafic rocks are composed of variably metamorphic and discontinuous Mg-and Fe-Ti gabbros (Fig. 1). MOR-type metabasalts cover a 2.5 km-long, N-S trending area within the sedimentary cover to the East of the Civrari serpentinites. Most of the Civrari ophiolitic rocks are overprinted by Alpine metamorphism, but locally fresh rocks are preserved.

Location and sample description
The Civrari peridotite consists of a very fresh, 10-15 m-wide lens of a clinopyroxene-and spinel-bearing harzburgite associated with cm-dm-thick gabbro dykes and occasional dm-thick serpentinised pyroxenites. The peridotites can be subdivided into refractory (sample PCB) and reactive harzburgites (sample PC2). The reactive peridotite displays a significant range in clinopyroxene shape and size, from rare 2 mm-wide coarse clinopyroxene (Fig. 2a) to rounded, anhedral and compositionally zoned 200-500 μm clinopyroxene rimmed by spinel or intergrowth of orthopyroxene + plagioclase (Fig. 2b). Clinopyroxene-spinel ± olivine intergrowths are found interstitially in both types of peridotites (Fig. 2c), whereas small (< 10 μm) apatite grains are found exclusively associated with plagioclase around partially dissolved spinel in the reactive peridotite (Fig. 2d). Plagioclase intergrown with orthopyroxene has been fully altered and replaced by a metamorphic assemblage of zeolites, Al-rich chlinochlore, Al-Mg hydroxides, garnet, clinozoisite, diopside and a variety of fine-grained hydrous aluminosilicates.
Basalts and gabbros were analyzed for whole-rock major and trace elements (Fig. 1b). Mg-gabbros are predominantly found at the contact with the Civrari serpentinite and are locally overlain by garnet-bearing quartzites. Isolated bodies of gabbros might be in direct contact with ophicarbonates and serpentinites or Fe-Ti gabbros. Mg-gabbros show locally preserved magmatic textures and preserved magmatic clinopyroxene. Five dm-cm-thick gabbro dykes intruding host peridotites, ranging from primitive olivine gabbros (MG-C1, Mu5, and Mu6) to slightly differentiated gabbros (Mu7, Mu8) were collected (Fig. 1a, b). These gabbros show well-preserved magmatic assemblages of clinopyroxene, plagioclase and olivine. Civrari gabbro MG-C1 is a composite gabbro with a troctolite zone composed of anhedral olivine and euhedral plagioclase grading abruptly into a gabbroic assemblage with olivine, euhedral plagioclase and subhedral to interstitial clinopyroxene (Fig. 2e, f). Lanzo gabbros show mm-cm large euhedral clinopyroxene, anhedral olivine, and interstitial plagioclase.

Analytical procedures
Whole-rock samples were powdered in an agate mill and dried to ca. 100 °C, with loss on ignition (LOI) determined by heating the sample to 1050 °C for 2 h. Li-tetraborate glasses were made by fusing 1.2 g of dried sample with 6 g of Lithium Tetraborate at 1300° for 3.5 min in Pt-crucibles followed by quenching. Major element compositions were measured on Li-tetraborate glasses and acquired by X-ray fluorescence using a Philips PW 2400 spectrometer at the Institute of Earth Sciences (ISTE), University of Lausanne (UNIL), Switzerland (Table 1). Cr and Zn for ultramafic rocks were measured by X-ray Fluorescence Philips PW 2400 spectrometer on compressed tablets of finely ground rock powder mixed with 20-30% of a cellulose wax. The standards SY-2, NIMN, NIMG, BHVO and UB-N were used as quality control (Govindaraju 1994). Uncertainties in XRF analyses are in the range 0.5 wt% (2σ) for major components such as SiO 2 to 0.01wt% for minor elements.  Fig. 1 a Lanzo-Civrari Ophiolite, Northern Italian Alps modified after Boudier (1978) and Pognante et al. (1986), with location of studied Southern Lanzo gabbros (GPS coordinates: N. 45°6′24, E. 7°28′62) as well as the Monte Civrari peridotites (GPS coordinates: N.45°11′32, E.7°19′53), the latter being represented by a red star. b Civrari Ophiolite, with location of samples. Oceanic sediments are Jurassic micaschists, calcschists, quartzites, and ophicarbonates In situ trace element abundances were measured on the flat side of broken off pieces of Li-tetraborate glasses using a quadrupole spectrometer Elan 6100 DRC for mafic rocks and a sector-field Inductively Coupled Plasma Mass Spectrometer ELEMENT2 XR interfaced to a NewWave UP-193 ArF excimer ablation system for ultramafic rocks. Analytical conditions were 10-20 Hz repetition rate and an energy of 160 mJ, which is equivalent to 12 J cm −2 with a beam size of 100 microns. The NIST SRM 612 glass standard was used to maximize sensitivity. Three repeat measurements were performed on each glass bead, with average standard deviation and detection limits listed in Table 2. Helium was used as a cell carrier gas. Dwell times for different isotopes range from 10 to 20 ms, employing a peak-hopping mode. Trace element concentrations were determined using CaO previously measured by XRF as an internal standard and NIST SRM 612 as an external showing only minor textural evidence of interaction with percolating MORB-melts. Ol olivine; b Rounded reactive clinopyroxene with orthopyroxene-plagioclase rims (opx and plg, respectively); c interstitial symplectite of Cr-spinel (spl) and clinopyroxene (cpx) between olivine in both reactive and refractory Civrari spinel-peridotites. d SEM image of a Cr-spinel with inclusions of 5-10 μm apatite grains (apt). The presence of apatite is highlighted in the red squares, which show P and Ca chemical maps. e Gabbroic section of the Civrari dyke, with plagioclase, clinopyroxene and olivine. f Troctolite section of the Civrari dyke, with olivine and euhedral plagioclase standard (Jochum et al. 2011). Data were processed using LAMTRACE software (Jackson 2008).
Representative mineral compositions were measured using a JEOL 8200 Superprobe (ISTE, UNIL) at 15nA, 15 kV with a spot size of 3 μm. The standards used were natural and synthetic silicates or oxides. Na and K were measured in first positions due to possible diffusion. Counting time was 16 s and 8 s on the peak and backgrounds, Per. Per.

Serp.
Serp.     respectively, for Na and K, and 30 s and 15 s for other elements. Measurements were corrected with the phi-rho-Z matrix correction method (Armstrong 1995). In situ trace element abundances were analyzed on a sector-field Inductively Coupled Plasma Mass Spectrometer ELEMENT2 XR interfaced to a NewWave UP-193 ArF excimer ablation system using a 100-75 μm spot size, 15-20 Hz repetition rate, equivalent to a 6 J/cm 2 on-sample energy density. Sensitivity was maximized using the NIST SRM 612 glass standard (La 139+ = > 2.5 × 10 6 c.p.s., Th 232+ = > 2.5 × 10 6 c.p.s.). The NIST SRM612 glass standard was used during analysis and SiO 2 or CaO, measured previously on the microprobe, were used as internal standards. Data were processed using LAMTRACE (Jackson 2008). Sr, Nd, and Pb isotope ratios were measured at the University of Geneva following the protocol and analytical conditions outlined in Chiaradia et al. (2011). Mineral separates were handpicked using a grain fraction between 250 and 100 μm. Mineral separates from samples Mu5, 6, and 7 were leached to remove any possible alteration following the method of Rampone et al. (1998). 100-200 mg of wholerock powders and mineral separates were leached to remove any secondary minerals with 3 M HCl prior to being rinsed with deionised water and dissolved in Teflon vials containing 4 ml of concentrated HF and 1 ml HNO 3 15 M at 140 °C on a hot plate for 7 days. Samples were then dried down and brought back up in 3 ml HNO 3 15 M prior to being dried down again. Sr, Nd, and Pb separation was carried out using cascade columns with Sr-spec, TRU-spec and Ln-spec resins following a modified method after Pin et al. (1994). Pb was further purified with a AG-MP1-M anion exchange resin in hydrobromic medium. Sr, Nd, and Pb isotope ratios were measured on a Thermo TRITON mass spectrometer on Faraday cups in static mode at the University of Geneva. Pb was loaded on Re filaments using the silica gel technique and all samples (and standards) were measured at a pyrometer-controlled temperature of 1450 °C. Pb isotope ratios were corrected for instrumental fractionation by a factor of 0.10% per amu based on more than 100 measurements of the SRM981 standard and using the standard values of Todt et al. (1996). External reproducibility of the standard ratios is 0.08% for 206 Pb/ 204 Pb, 0.12% for 207 Pb/ 204 Pb and 0.16% for 208 Pb/ 204 Pb. Sr was loaded on single Re filaments with a Ta oxide solution and measured at a pyrometer-controlled temperature of 1480 °C in static mode using the virtual amplifier design to cancel out biases in gain calibration among amplifiers. 87 Sr/ 86 Sr values were internally corrected for fractionation using a 88 Sr/ 86 Sr value of 8.375209. Raw values were further corrected for external fractionation by a value of + 0.03‰, determined by repeated measurements of the SRM987 standard ( 87 Sr/ 86 Sr = 0.710250). External reproducibility of the 87 Sr/ 86 Sr ratio for the SRM987 standard is 7 ppm. Nd was loaded on double Re filaments with 1 M HNO 3 and measured in static mode with the virtual amplifier design. 143 Nd/ 144 Nd values were internally corrected for fractionation using a 146 Nd/ 144 Nd value of 0.7219 and the 144 Sm interference on 144 Nd was monitored on the mass 147 Sm and corrected using a 144 Sm/ 147 Sm value of 0.206700. Raw values were further corrected for external fractionation by a value of + 0.03‰, determined by repeated measurements of the JNdi-1 standard ( 143 Nd/ 144 Nd = 0.512115: Tanaka et al. 2000). External reproducibility of the JNdi-1 standard is < 5 ppm.

Bulk rock trace elements
REE patterns of ultramafic rocks (Fig. 5) display a strong fractionation in Light-REE (La N /Yb N = 0.017-0.23) and Middle-REE (Sm N /Yb N = 0.07-0.22) (N = normalized to Chondrite-C1, after McDonough and Sun 1995). All samples show fractionated REE patterns (La N /Sm N between 0.2 and 1.1) and Heavy-REE at or below Chondrite values, showing refractory compositions relative to most Western Tethyan mantle rocks (e.g., Müntener et al. 2010, Kaczmarek and. Serpentinites show Ce, U, and Pb enrichment as well as varying Ba, Rb, and Eu abundances, consistent with seawater serpentinisation of the host peridotite (e.g., Hart 1973, Burgath et al. 1997) (see electronic supplementary material ESM1, Fig. 1).

Mineral chemistry
The complete data set can be found in the electronic supplementary material (ESM2, Tables 1, 2, 3, 4).

Clinopyroxene
Clinopyroxene is found as euhedral to anhedral grains. Orthopyroxene exsolution lamellae ≤ 5 micrometers along cleavage planes thin out at the rims. Interstitial clinopyroxene is intergrown with spinel, forming symplectitic textures. Textural observations indicate that clinopyroxene grain size varies more strongly in the sample with reaction textures, leading to smaller, rounder grains with  (Niu et al. 2002) and Mid-Atlantic Ridge (MAR) MORB from Bryan et al. (1981). 1 Gpa and 0.7 Gpa crystal fractionation trends for major elements of dry tholeiitic basalts from Villiger et al. (2004Villiger et al. ( , 2007, and 1 atm trend follows 1 atm experimental data from the literature and compiled in Villiger et al. (2007) rims of orthopyroxene-plagioclase intergrowths (Fig. 2a,  b).
Trace element compositions of clinopyroxene range from homogeneous and strongly depleted (Ce N / Yb N = ~ 0.001), consistent with depleted clinopyroxene from abyssal peridotites, to light-REE enriched with core-rim zonations (Ce N /Yb N = 0.037-0.11) (Fig. 7a,  b). Reacted clinopyroxene surrounded with orthopyroxene + plagioclase intergrowth display the highest Light-REE. Some reacted clinopyroxene cores show similar trace element patterns (e.g., Zr/Hf and Middle-REE) and major elements compositions as non-reacted refractory clinopyroxene (Fig. 7a, b). Abundances in Ce, Ti, Na 2 O, and Zr in clinopyroxene increase from residual to reactive peridotites (Fig. 8). In contrast, compatible elements such as NiO decrease in the reactive peridotite (Fig. 8). Zr/Hf ratios are low and near-homogenous for residual peridotites (2-3.2) but significantly higher and more heterogeneous for reacted clinopyroxene (5.6-60.7).
Trace element patterns of the Civrari gabbro clinopyroxene are more Light-REE depleted (La N /Sm N = 0.078 and La N /Yb N = 0.061) than primitive Lanzo gabbro clinopyroxene (La N /Sm N = 0.122 and La N /Yb N = 0.103). More evolved gabbros show an increase in REE and increasing negative Eu anomaly in clinopyroxene (Fig. 10a, b). Interstitial clinopyroxene in evolved Lanzo gabbros is enriched in REE, incompatible trace elements, and a pronounced depletion in Eu, consistent with crystallization of trapped melt fractions in equilibrium with plagioclase.

Plagioclase
Euhedral plagioclase from the Civrari gabbro ranges from An75 to An69, while the Lanzo gabbros have more evolved plagioclase, ranging from An64 to An49. Civrari plagioclase displays flat Light-REE pattern (La N /Sm N between 1 and 1.2) and Lanzo gabbros show Light-REE enriched patterns, with La N /Sm N between 1.7-2.0 and 3.1-4.5 for primitive and evolved gabbros, respectively (Fig. 10c).

Olivine
Olivine is anhedral and mostly altered. Civrari gabbro olivine is homogenous and have a Mg# between 86.2 and 86.7 and NiO concentration averaging 0.22 wt%. Table 4 Geothermometry of Civrari residual peridotites T Ca and T BKN are the Ca-in-orthopyroxene thermometer and the two pyroxene thermometer of Brey and Köhler (1990), respectively; T WE-S is the clinopyroxene-orthopyroxene thermometer of Witt-Eickschen and Seck (1991); T W-W is the Webb and Wood (1986) thermometer based on the Cr/Al relationship of spinel and pyroxenes; T V , T Sc , and T Co are pyroxene trace element thermometers from Seitz et al. (1999); and T REE is the pyroxene REE thermometer from Liang et al. (2013) )  1120  1120  1130  1240  1240  1320  1310  1380  1σ  75  35  63  35  18  27  35 Parkinson et al. (2003) with compositional fields from Bonatti and Michael (1989), showing the compositional evolution of spinel and orthopyroxene during partial melting. b Cr# spinel vs Cr# clinopyroxene. Dark grey area: spinel and plagioclase peridotites from the Eastern Central Alps . Light grey area, dotted rim: spinel-peridotites from the Lena through (Hellebrand and Snow 2003). Trend A: compositional evolution of orthopyroxene, clinopyroxene and spinel during melt percolation

Residual nature of Civrari peridotites and ancient melting events
Low bulk Al 2 O 3 and TiO 2 abundances highlight the largescale homogeneity and residual nature of the Civrari ultramafic massif (Fig. 3). This contrasts sharply with the compositional heterogeneity of Western Tethys peridotites from the European Alps and Apennines and more closely resembles the compositional spectrum of residual abyssal peridotites. The depleted bulk composition of Civrari serpentinites and peridotites (Figs. 3, 5a, b) is the result of fractional melting. The Cr# and TiO 2 abundances of spinel and low Na 2 O contents and strongly fractionated REE trends in clinopyroxene (Figs. 7,8) are consistent with a residual signature after ca. 3% melting in the garnet stability field and ca. 10-11% melting in the spinel stability field (McCarthy and Müntener 2015). Local microtextures of interstitial clinopyroxene-spinel ± olivine symplectites Note the rapid enrichment in Light-REE and Na prior to Ti enrichment and Ni depletion. d Zr (ppm) vs Ti (ppm) of mantle clinopyroxene during cryptic melt percolation; e Zr (ppm) vs Zr/Hf ratio of mantle clinopyroxene. Fractional melting modeling using starting modes and melting modes of Kinzler (1997), composition of DMM clinopyroxene from Workman and Hart (2005) and fractional melting equation from Johnson et al. (1990). Partition coefficients are from Suhr et al. (1998) and Hart and Dunn (1993), Kelemen et al. (1993) and Kelemen et al. (2003). Data for comparison: continental xenoliths are from Weyer et al. (2003), and Downes et al. (2003Downes et al. ( , 2015. Southwest Indian Ridge (SWIR) clinopyroxene and classification (residual, depleted, intermediate, and enriched) from Warren and Shimizu (2010). Field of Western Tethys clinopyroxene (spinel and plagioclase peridotites) from Müntener et al. (2004), Tribuzio et al. (2004), Rampone et al. (2008) and Müntener et al. (2010). Small grey circles correspond to the compilation of abyssal peridotites from Warren (2016) and Mid-Atlantic Ridge peridotites from Brunelli et al. (2006) indicating low-volume melt retention in the stability field of spinel. Whether or not the Civrari peridotites represent residues of near-fractional melting related to the spatially associated basalts and gabbros remains controversial. Several residual spinel-peridotites from the Western Tethyan mantle are not in isotopic equilibrium with spatially associated lavas (e.g., Rampone et al. 1998;Müntener et al. 2004), as also illustrated by Civrari peridotites and associated magmatism (Fig. 11b, c). One hypothesis is that the Tethyan upper mantle is isotopically heterogeneous, which leads to a geochemically heterogeneous oceanic lithosphere (e.g., Cipriani et al. 2004;Sanfilippo et al. 2019). This hypothesis is supported by mantle heterogeneity along modern midocean ridge systems (e.g., Stracke et al. 2011;Warren et al. 2009). An alternative hypothesis suggests that mantle heterogeneity is inherited from the subcontinental lithosphere (McCarthy and Müntener 2015;Rampone et al. 1998;Müntener et al. 2004). Indeed, parts of the western Tethyan ophiolites are derived from former ocean-continent transition zones (e.g., Müntener et al. 2010;Manatschal & Müntener 2009). Thus, rift-related extension leads to the exhumation of subcontinental mantle with inherited ancient melting and metasomatic imprints, locally erased by refertilisation during incipient opening of the Western Tethys (e.g., Picazo et al. 2016). This alternative interpretation is also supported by the presence of cold (850-1050 °C) spinel-harzburgites affected by high-degree partial melting (14-25%) along the magma-poor Newfoundland ocean-continent transition zone (Müntener and Manatschal 2006). Depleted spinelperidotites, found throughout the Alpine-Apennine orogen show a coherent Nd-isotopic "errorchron" at 273 ± 24 Ma (McCarthy and Müntener 2015) (Fig. 11b). This isotopic trend over a large geographical region is surprising but consistent with widespread Permian-age mafic magmatism throughout Western Europe (e.g., Pin 1986;Schaltegger and Brack 2007). This might be a coincidence, but supports the alternative interpretation that isotopic disequilibrium in western Tethys ophiolites is related to inherited, ancient melting events.
Despite evidence for melt metasomatic processes in the subcontinental mantle worldwide (Wiechert et al. 1997;Witt-Eickschen and Kramm 1998;Downes 2001;Ionov et al. 2002), the homogeneous composition of Civrari peridotites and serpentinites indicates that km-scale subcontinental mantle domains may remain unaffected by metasomatic processes for over 100 Ma. Depleted characteristics of peridotites in (ultra)slow-spreading systems therefore cannot be used to discount a subcontinental mantle origin (e.g., Lassiter et al. 2014).

Rapid cooling of high-temperature Civrari peridotites
Equilibrium temperatures for the refractory spinel peridotite at 1.5 GPa are reported in Table 4. Geothermometers based on major elements have lower calculated temperatures (1120-1240 °C) than thermometers calculated using trace elements (1240-1380 °C) ( Table 4). The T REE of Liang et al. (2013), based on slow diffusion elements, gives the highest calculated T°, reaching 1380 ± 47 °C, which is within the range of the dry peridotite solidus of ca. 1300-1350 °C at 1.5-2 GPa (Hirschmann 2000). These elevated temperatures are consistent with elevated CaO in orthopyroxene (~ 2 wt%), favouring the interpretation that Civrari peridotites represent a rare case of near-fractional melting (Fig. 7a). These elevated temperatures are unlike those estimated for other Western Tethys mantle peridotites, which range from  Rampone et al. (1998), Desmurs et al. (2002) and Montanini et al. (2008). Gab-bro clinopyroxene are from the Southwest Indian Ridge (SWIR) and the Mid-Atlantic Ridge (MAR) (Coogan et al. 2001;Niu et al. 2002;Ross and Elthon 1997) 69 Page 18 of 25 850 to 1100 °C (Rampone et al. 1996;Rampone et al. 2005;Müntener et al. 2010). The comparison of T BKN and T REE indicates fast cooling rates similar to abyssal peridotites and significantly faster than cooling rates recorded by subcontinental mantle peridotites along the Western Tethys (Dygert and Liang 2015). The Civrari peridotites thus represent either rapid cooling of an abyssal peridotite at an active spreading ridge, or, alternatively, implies rapid cooling of subcontinental lithosphere. As discussed previously, we interpret the isotopic decoupling between western Tethyan mantle and mafic rocks as caused by ancient (273 ± 24 Ma) near-fractional melting unrelated to Jurassic MOR-type magmatism (Fig. 11b, c) Fig. 11 a Present-day 87 Sr/ 86 Sr vs 143 Nd/ 144 Nd of Civrari mafic rocks, clinopyroxene, and plagioclase of Civrari and Lanzo gabbros.; b 147 Sm/ 144 Nd vs present-day 143 Nd/ 144 Nd of Lanzo and Civrari gabbros. Black circles represent residual Alpine spinel-peridotites from the Western Tethys (data from Tribuzio et al. 2004;Rampone et al. 1996;Müntener et al. 2004). c 147 Sm/ 144 Nd vs age-corrected (160 Ma) 143 Nd/ 144 Nd of Lanzo and Civrari gabbros compared to the Civrari residual peridotites (data from McCarthy and Müntener 2015). The 160 Ma age corresponds to the age of MORB magmatism during opening of the Western Tethys (Piemont-Liguria ocean) (e.g., Tribuzio et al. 2016;. Tethyan MORB includes gabbros and basalts from the Internal Liguride (Rampone et al. 1998), Platta (Schaltegger et al. 2002), External Liguride (Tribuzio et al. 2004) and Montgenèvre Ophiolite (Costa and Caby 2001). MORB basalts are from White and Hofmann (1982) and Ito et al. (1987) Page 19 of 25 69 and rapid cooling as a result of rapid exhumation from a deeper, warmer section of the subcontinental mantle (e.g., Picazo et al. 2016).

Low-pressure MORB crystallization in exhumed subcontinental mantle
The most enriched mantle clinopyroxene evolve towards the compositional field of clinopyroxene from gabbroic dykes (Figs. 7,8). This is illustrated by the enrichment of Ce, Ti, Zr and Na 2 O as well as decreasing Ni of reactive mantle clinopyroxene towards the composition of gabbroic clinopyroxene (Fig. 8). We argue that the melt percolating through and reacting with the Civrari peridotite was of similar composition as the spatially associated gabbros. Pb isotopic composition (see electronic supplementary appendix ESM1, Fig. 3), unradiogenic 87 Sr/ 86 Sr (< 0.703), and radiogenic Nd (εNd = 7-10) of unaltered whole-rocks and mineral separates indicate MORB-like isotopic compositions (Fig. 11). This is also consistent with the bulk composition of Civrari basalts, which are similar to mid-ocean ridge basalts (Fig. 4). Clinopyroxene compositions from gabbros and reactive clinopyroxene resemble clinopyroxene from midocean ridge gabbros (e.g., Coogan et al. 2001) (Figs. 9, 10a, b). Thus, isotopic compositions, mineral chemistry and bulk rock chemistry indicate that ascent of MORB-melts leads to the emplacement of gabbroic dykes and localized melt percolation during cooling of the peridotites.
The compositional variation of Civrari gabbros mirrors the range found in other Tethyan ophiolites Montanini et al. 2008) (Fig. 4). Most Mg-gabbros have positive Eu, Sr anomalies and high Al 2 O 3 contents suggesting crystal fractionation dominated by plagioclase whereas gabbros with lower Al 2 O 3 and increasing Cr and Ni contents (Fig. 4) suggest accumulation of clinopyroxene (± spinel) and olivine, respectively. Bulk rock gabbro compositions show compositional trends similar to fractionating MORBs (Figs. 4,5b), with increasing Zr, P 2 O 5 , and TiO 2 with decreasing Mg#. Textures from the Civrari gabbro dyke indicate a crystallization sequence of olivine > plagioclase > clinopyroxene, consistent with low-pressure evolution of MORB melts (< 5kbar) (Grove et al. 1993). Olivine, clinopyroxene and plagioclase compositions are comparable to primitive gabbros crystallized at 3-5 kbar in the Mid-Cayman Rise Ross and Elthon 1997) (Fig. 9). Jurassic gabbros and basalts from the Western Tethys plot along a low-pressure fractionation trend favouring enrichment of TiO 2 and FeO in MORB (e.g., Villiger et al. 2007), leading to late-stage accumulation of ilmenite, titanomagnetite, apatite, and the formation of evolved cumulates with high TiO 2 and P 2 O 5 (Fig. 4).
Euhedral clinopyroxene from Lanzo gabbros has higher Mg# and Al 2 O 3 , suggesting slightly higher pressure of crystallization (ca. 5kbar) (Grove et al. 1993;Langmuir et al. 1992;Feig et al. 2006) (Fig. 9b). Increasing pressure will lead to a suppression of plagioclase and olivine primary phase fields and an expansion of the clinopyroxene field (Presnall et al. 1978;Sen and Presnall 1984), leading to clinopyroxene with higher Al 2 O 3 and Cr 2 O 3 and less-pronounced negative Eu anomalies as well as higher NiO. Primitive Lanzo clinopyroxene shows less fractionated Light-REE and higher Na 2 O abundances as well as higher La/Sm ratios in plagioclase compared to Civrari gabbro clinopyroxene and plagioclase (Figs. 9, 10). Similar isotopic compositions (Fig. 11) indicate that the compositional variability of Civrari and Lanzo gabbros results primarily from slight differences in the degree of partial melting of a depleted mantle source.

Cryptic grain-scale infiltration of MORB-melts
The reactive harzburgite shows distinct microtextures indicative of melt-rock interaction coupled with high MgO and low SiO 2 implying high olivine abundances. Primary MORB-type melts originating from partial melting of peridotite at 10-30 kb will be saturated in olivine + Al-phase (Kelemen 1990) leading to pyroxene dissolution and olivine crystallization, forming reactive high-MgO mantle rocks. Sisaturation of ascending melt will then lead to the dissolution of clinopyroxene and crystallization of orthopyroxene + plagioclase in the plagioclase stability field, consistent with the presence of orthopyroxene + plagioclase intergrowth around rounded, anhedral mantle clinopyroxene and spinel grains (Fig. 2b). TiO 2 in spinel is correlated with increasing Cr#, consistent with reaction of a MORB-melt and crystallization of plagioclase (Dick 1989). However, even though reacted clinopyroxene show higher Na 2 O concentrations and higher Cr# in spinel, consistent with equilibrium with plagioclase (e.g., Müntener et al. 2010), they retain a refractory composition compared to other clinopyroxene from Western Tethyan ultramafic rocks (Fig. 8). This indicates that melt percolation at a local-scale is not as pervasive as generally recorded in Western Tethys ophiolites (Piccardo et al. 2007;Müntener et al. 2010). On the scale of a thin sections (2.2 × 3.7 cm), reactive clinopyroxene varies from mmsize subhedral grains to rounded ~ 150 μm grains rimmed by orthopyroxene-plagioclase intergrowth ± apatite (Fig. 2d), indicating localized melt percolation and crystallization. The lack of equilibration of the peridotite with ascending melts is indicated by core-rim zonation in major and trace elements in mantle pyroxenes (Figs. 6,7,8). The presence of 3-5 μm apatite suggests trapped melt crystallization leading to apatite saturation due to increasing P 2 O 5 in residual melt fractions. Fractional crystallization calculations at 4 kbar using Rhyolite-MELTS (Gualda et al. 2012) with a primitive basalt HK #19 (Hirose and Kushiro 1993) and 0.08 wt% P 2 O 5 and 0.1 wt% H 2 O indicates that apatite saturation in MORB-type melts occurs just below 1100 °C, which provides an upper bound for ambient mantle temperature during melt entrapment.
Grain-scale melt percolation lead to well developped core-rim zonation patterns in clinopyroxene (Figs. 7, 8). The enrichment trend of reacted clinopyroxene overlaps the field of both residual and reactive clinopyroxene from abyssal peridotites (Warren and Shimizu 2010) (Fig. 8) and illustrates how sensitive Na 2 O and Light-REE are to local interactions with percolating melts. As can be seen in peridotites from the SWIR identified as residues after partial melting (Warren and Shimizu 2010) Na 2 O varies by a factor of five with no concomitant change in Ce (Fig. 8b). This indicates that these residual abyssal peridotites have been affected by localized, small-scale melt percolation or melt retention (e.g., Warren 2016).
Abyssal peridotites showing elevated Cr# in spinel, elevated Na 2 O abundance and enrichement in light-REE compared to Civrari residual peridotite cannot be readily explained by near-fractional melting. These chemical trends offer further support indicating that most residual abyssal peridotites show weak refertilisation and/or that most peridotites are not solely a result of near-fractional melting but retain significant amounts of melt upon decompression (Elthon 1992;Hellebrand et al. 2002;Brunelli et al. 2006;Warren and Shimizu 2010;Warren 2016). This demonstrates the importance of Na 2 O concentrations in clinopyroxene prior to determining their residual character, as clinopyroxene is prone to rapidly reequilibrate with diffuse, lowvolume melts (e.g., Lundstrom 2000).

Fractionating Zr-Hf ratios in mantle clinopyroxene: near-fractional melting and cryptic MORB-melt percolation
Civrari mantle clinopyroxene shows strongly variable Zr/ Hf ratios (1-60) (Fig. 8e). These variations contrast with the proposal that Zr/Hf ratios are uniform and chondritic (Jochum et al. 1986). The Zr/Hf ratios of Civrari residual and refractory clinopyroxene overlap with the Zr/Hf ratios of subcontinental mantle xenoliths (Fig. 8e) (Lenoir et al. 2000;Downes et al. 2015). Civrari clinopyroxene show low Zr/Hf ranging between 2.1 and 3.2 and are the lowest ratios recorded in Western Tethys ophiolites (e.g., Müntener et al. 2010) and much lower than Chondrite (ca. 34.2, Jochum et al. 1986). These values are consistent with fractional melting modeling in the spinel stability field of an initial Depleted MORB Mantle (DMM) (Workman and Hart 2005) (Fig. 8e). Strongly subchondritic to superchondritic Zr/Hf (e.g., 0.5-82), high-depletion of Zr and Hf and elevated εHf at moderate εNd have been documented in xenoliths derived from the subcontinental mantle of Western Europe (Wittig et al. 2006;Downes et al. 2015). To the contrary of those samples that show enriched Light-REE implying mantle metasomatic processes, the Civrari residual peridotites were not affected by metasomatism. This implies that Zr/Hf fractionation results from near-fractional melting as a consequence of a slight difference in compatibility of Zr and Hf during mantle melting (e.g., David et al. 2000;Weyer et al. 2003;Stracke et al. 2011).
Cryptic percolation of MORB-melt in the reactive peridotite leads to significant variations in Zr/Hf ratios (20-60) at near-constant Zr at the grain scale (Fig. 8e). Zr/Hf ratios are significantly more variable, and Zr generally lower, than coeval clinopyroxene from intruding gabbros and from refertilised plagioclase peridotites ). As reacted clinopyroxene show enrichment trends evolving towards the composition of coeval MORB gabbro clinopyroxene (e.g., Figs. 7a, 8c) the variation in Zr/Hf in the case Civrari peridotites cannot be related to reaction of peridotites with exotic (e.g., carbonatite) melts, as proposed by Dupuy et al. (1992), Rudnick et al. (1993) and Downes et al. (2003) for intra-plate magmatism and metasomatised mantle xenoliths. The main distinction between refertilised plagioclase peridotites and reactive Civrari peridotites is related to the significant change in melt/rock ratio upon melt percolation. Müntener et al. (2010) calculated that up to 12% MORB-melt is retained during pervasive refertilization. This elevated melt/rock ratio is consistent with the Zr/Hf of these clinopyroxene overlapping the composition of gabbroic clinopyroxene (Fig. 8e). On the contrary, the textural and compositional variation of Civrari reactive clinopyroxene at the grain scale implies significantly lower melt/rock ratios. Thus, percolation of MORB-type melts at low melt/ rock ratio fractionates Zr/Hf as a result of slight differences in compatibiliy of Zr and Hf (e.g., Navon and Stoler 1987;Bodinier et al. 1989). Cryptic percolation of MORB-melts upon rifting and mantle exhumation in ocean-continent transitions might lead to compositional variations in trace elements crucial for the long-term compositional evolution of the mantle lithosphere.
Clinopyroxene with elevated radiogenic Hf signatures (εHf 160Ma : 14-213) at MORB-like εNd (6-15) have been described in reactive harzburgites in western Tethys ophiolites (Southern Lanzo Massif, Sanfilippo et al. 2019). One possibility is that this isotopic signature is reflecting melting of "ultra-depleted" asthenospheric domains. However, western Tethys ophiolites have significant isotopic and petrological heterogeneities at small and large scale (e.g., Rampone and Hofmann 2012;Piccardo et al. 2007). Therefore, an alternative is that MORB-melts might have percolated into heterogeneous, subcontinental mantle with radiogenic Hf-Nd isotopic signatures inherited from previous melting and metasomatic events. MORB-melt percolation would equilibrate Nd isotopic ratios to MORB values more efficiently than Hf isotopic ratios ). The analysis of Hf-isotopic ratios of depleted western Tethys spinel-peridotites unaffected by subsequent melt percolation would allow to test these competing hypothesis. For example, assuming that the Civrari peridotite was in equilibrium with a DMM at 270 Ma upon partial melting (McCarthy and Müntener 2015), we predict, using the average and 2σ of Lu and Hf abundances of Civrari clinopyroxene and known DMM compositions (Workman and Hart 2005;Salters and Stracke 2004), that the measured 176 Hf/ 177 Hf should be within a 2σ range of εHf = 81-121 (average of εHf = 94-96). This implies a εHf 160Ma = 47-64 (average of 52-54), distinctly different than the εNd 160Ma of 19-20 (McCarthy and Müntener 2015). Such elevated εHf 160Ma abundances would be well within the range of εHf of mantle peridotites from the Gakkel Ridge, mantle xenoliths and reactive harzburgites from the western Tethys Byerly and Lassiter 2014;Sanfilippo et al. 2019;Downes et al. 2015) and significantly more radiogenic than MORB (Fig. 12).

Conclusions
The Civrari peridotite shows similar petrological and geochemical characteristics as residual abyssal peridotites. Mantle clinopyroxene shows some of the most fractionated REE and lowest Na 2 O ever measured in abyssal peridotites and imply that these peridotites are the product of near-fractional melting starting in the garnet stability field. Melt percolation of MORB-melts similar in composition to coeval gabbroic intrusions lead to grain scale melt-rock interaction and induced the compositional zonation of reacted mantle clinopyroxene. We show that diffuse-melt percolation affects Na 2 O abundances of mantle clinopyroxene. The high T° (1200-1300 °C) recorded by these peridotites and the preservation of grain-scale compositional zonations are the consequence of rapid cooling and exhumation to the seafloor during Jurassic extension. The lack of isotopic equilibrium between mafic rocks and mantle peridotite suggests that partial melting is ancient and unrelated to MORB-type magmatism. Therefore, though Civrari peridotites are compositionally unlike other mantle rocks from the Western Tethys and show key characteristics of abyssal peridotites, such as km-scale homogeneity, a depleted nature, elevated recorded temperatures and fractional melting with limited melt percolation, they are the product of protracted magmatic processes occurring during Jurassic rifting superimposed over ancient (Permian) melting events.  ) and mantle xenoliths . Lanzo peridotites (western Tethys) are from Sanfilippo et al. (2019). Predicted field of εHf of Civrari is in orange; εNd values are from McCarthy and Müntener (2015). Note that these are present-day values, except for Civrari peridotites and Lanzo peridotites, which have been corrected to 160 Ma