Abiotic passive nitrogen and methane enrichment during exhumation of subducted rocks: Primary multiphase fluid inclusions in high‐pressure rocks from the Cabo Ortegal Complex, NW Spain

Primary multiphase fluid inclusions (MFI) were studied in one eclogite and two granulites from the Cabo Ortegal Complex (COC, NW Spain) by means of Raman imaging, Scanning Electron Microscopy with Energy Dispersive Spectroscopy (SEM‐EDS) and Focused Ion Beam ‐ Scanning Electron Microscopy (FIB)‐SEM. Complementary, secondary MFI in pyroxenites from COC were also investigated. MFI hosted in eclogite and granulites occur along growth zones or in 3D clusters in garnet porphyroblasts suggesting a primary origin at high‐pressure (HP) metamorphic conditions. The mineral assemblage of MFI is mainly composed of Fe‐Mg‐Ca‐carbonates and phyllosilicates ± graphite ± quartz ± corundum ± pyrite ± apatite ± rutile and a fluid phase composed of nitrogen ± methane ± carbon‐dioxide. The mineral proportions vary among the lithologies. Dominant carbonates and hydrous silicates are interpreted as step‐daughter minerals (crystals formed in the MFI after entrapment as a result of fluid–host interaction), whereas apatite, quartz and rutile are considered in part as accidentally trapped minerals since they also occur as crystal inclusions together with MFI in each rock type. Quartz and corundum occur together in MFI in ultramafic granulite and are regarded as step‐daughter minerals in this lithology. These observations suggest that the MFI are products of post‐entrapment reactions of a homogeneous COHN fluid system with the host mineral. Thermodynamic calculations in the CaFMAS‐COHN system confirmed that bulk composition of the MFI in eclogite is similar to the host garnet+COHN composition except for a potential lost of H2O. Carbonation and hydration reaction between the host (i.e. garnet or pyroxene) and the fluid inclusion results in the consumption of all CO2 and part of the H2O from the fluid phase producing Ca‐Fe‐Mg‐carbonates and hydrous step‐daughter minerals, mostly pyrophyllite and chlorite. Nitrogen content of the originally trapped COHN fluid in eclogite was estimated to have a maximum value of 10 mol% at peak HP conditions and 30–40 mol% at retrograde conditions that is within the range of the observed MFI in the residual fluid (13–68 mol%). Pseudosection modelling confirmed the stability of the phase assemblage in the MFI in a specific low‐pressure, low‐temperature stability field (between 300°C and 400°C at pressures < 1 GPa), caused by H2O‐ and CO2‐consuming reactions possibly in a single step. Our findings indicate that such processes in the exhuming HP units may play a role in global nitrogen and carbon cycling as well as potentially contributing to nitrogen and methane supply to subsurface–surface environments during devolatilization in the forearc regions of convergent plate margins.

trapped COHN fluid in eclogite was estimated to have a maximum value of 10 mol% at peak HP conditions and 30-40 mol% at retrograde conditions that is within the range of the observed MFI in the residual fluid (13-68 mol%).
Pseudosection modelling confirmed the stability of the phase assemblage in the MFI in a specific low-pressure, low-temperature stability field (between 300 C and 400 C at pressures < 1 GPa), caused by H 2 O-and CO 2 -consuming reactions possibly in a single step. Our findings indicate that such processes in the exhuming HP units may play a role in global nitrogen and carbon cycling as well as potentially contributing to nitrogen and methane supply to subsurface-surface environments during devolatilization in the forearc regions of convergent plate margins.
Nitrogen species preserved in rocks provide a direct record on past geochemical nitrogen cycling. The balance of its input/output during global recycling processes has been widely investigated, mostly by isotopic studies (e.g. Busigny & Bebout, 2013;Halama et al., 2012). Volatile recycling by conveying subducted ultramafic units evolved along the slab-mantle interface provide important contributions to the deep Earth diamond genesis (N may substitute C in concentrations up to 11,000 ppm) (Cannaò et al., 2020;Cartigny et al., 2001;Cartigny & Marty, 2013;Halama et al., 2012). Studying the nature of nitrogen during its transport and retention into deep reservoirs together with degassing at convergent margins is of great importance for our understanding of the planetary atmospheric evolution (Mikhail & Sverjensky, 2014). Although subduction zone cycling defines nitrogen transfer together with non-radiogenic noble gas isotopes as the major exchange between near-surface geospheres and deep Earth reservoirs (Kendrick et al., 2011(Kendrick et al., , 2018, the mechanism governing the enrichment of nitrogen and explanation of its abundance in forearc regions of accretionary subduction complexes is poorly understood. The Cabo Ortegal Complex (COC) in NW Spain is formed by metamorphic rocks from a former subduction zone and is a well-studied locality in terms of petrologic, geochemical and structural aspects (e.g. Ábalos et al., 1996, 2003Albert et al., 2015;Gil Ibarguchi et al., 1990, 1999Henry et al., 2017;Llana-Fúnez & Brown, 2012;Marcos et al., 2002;Ord oñez Casado et al., 2001;Puelles et al., 2005Puelles et al., , 2009Tilhac et al., 2016Tilhac et al., , 2017. By contrast, the fluids, which were present during different stages of its pressure and temperature (P-T) evolution, are poorly constrained. In this study, we characterize for the first time and then discuss the significance of multiphase fluid inclusions (MFI) preserved in metamorphic rocks from the COC. We propose a possible evolution of such trapped fluids during the retrograde exhumation path supported by thermodynamic modelling. The passive enrichment of nitrogen during post-entrapment carbonation and hydration processes with their hosts observed in MFI provides new constraints to better understand the preservation of subduction fluids and nitrogen cycling in subduction zones.

| GEOLOGICAL BACKGROUND
The COC, located in the northwestern edge of the Iberian Peninsula, forms the upper structural unit of the Allochtonous Complexes exposed in the Iberian Massif ( Figure 1) (Ábalos et al., 2003;Albert et al., 2015;Martínez Catal an et al., 2009). These complexes belong to the Variscan belt of Europe and are interpreted as fragments of continental and oceanic lithosphere, which were involved in a Devonian subduction-exhumation cycle (Puelles et al., 2005). The rocks of the COC record a multistage tectonothermal history as a result of prograde and retrograde metamorphism. The structural units, separated by ductile thrust contacts, are thought to have been amalgamated and undergone HP deformation in deep tectonic realms of a Variscan subduction/collision orogenic channel (Ábalos et al., 2003;Puelles et al., 2005). The studied area comprises high-pressure and high-temperature (HP-HT) ultramafic massifs associated with eclogites, granulites, orthogneisses, paragneisses and metagabbros ( Figure 1) (Arenas, 1991; F I G U R E 1 Simplified geological map of the Cabo Ortegal Complex (COC) modified afterÁbalos et al. (2010). Locations of the sampled and selected eclogites, granulites and pyroxenites are indicated with colored stars; sample names are indicated for selected samples in this study Den Tex, 1961;Gil Ibarguchi et al., 1990, 1999Vogel, 1967). The origin of the main HP metamorphic units is rather well constrained: subcontinental lithospheric mantle for the peridotites and pyroxenites, middle to upper crustal volcanic arc for the granulites, N-MORB for the eclogites and shallow crustal volcanosedimentary sequences for the gneisses (Ábalos et al., 2003;Santos Zalduegui et al., 2002). Pyroxenites are thought to have been formed by melt-peridotite interaction in a suprasubduction setting around 459-515 Ma ago based on Rb-Sr and Sm-Nd dating (Santos Zalduegui et al., 2002;Tilhac et al., 2017). The following PT conditions of peak metamorphism have been acquired for the main units of the COC: ultramafics (1.6-1.8 GPa, 780-800 C), eclogite (1.8-2.2 GPa, 700-800 C), granulite (1.4-1.7 GPa, 740-835 C) and gneiss (approximately 1.5 GPa, 700 C) (Ábalos et al., 2003;Gil Ibarguchi et al., 1990;Mendia, 2000;Puelles et al., 2005;Tilhac et al., 2016Tilhac et al., , 2017. The age of peak metamorphism (approximately 382-389 Ma U-Pb on titanite and zircon; SHRIMP U-Pb on zircon and internal Sm-Nd isochrons on Cpx-Grtwhole-rock), recorded in eclogites and granulites, corresponds to the formation and accretion of the HP sheets of the units of the COC at sub-Moho depths (Ábalos et al., 2003;Ord oñez Casado et al., 2001;Peucat et al., 1990;Santos Zalduegui et al., 2002). A single PT-loop has been assigned for the metamorphic evolution of granulites as well during Variscan subduction, as being a part of a coherent ensemble together with other HP rocks in the COC during exhumation (Puelles et al., 2005). Simultaneously, ultramafic rocks have undergone eclogite facies metamorphism and deformation. This was followed by percolation of hydrous fluids causing metamorphic amphibolization of pyroxenites and hydration of their host peridotites . The presence of a low-viscosity subduction channel, active between ca. 395 and 365 Ma, and fast exhumation of HP units with a rate comparable to plate motions has been inferred for the serpentinite-matrix mélange rocks, which occur in the lowest structural position of the COC (Ábalos et al., 2003, 2010Novo-Fern andez et al., 2016).
The eclogitic unit of the COC forms one of the largest known continuous eclogite outcrops with a ridge 100-700 m thick and 17 km long in the direction N-NE to S-SW (Ábalos, 1997). Detailed petrographic, mineralogical, geochemical, structural and thermobarometric data of eclogites are provided byÁbalos et al. (2010), Gil , Mendia (2000) and Vogel (1967). Based on the consistency of isotopic and geochemical characteristics, the whole eclogite massif has an oceanic crustal origin (Bernard-Griffiths et al., 1985;Peucat et al., 1990), representing different stages of N-MORB tholeiitic series with variable degrees of differentiation (Ábalos et al., 2010;Mendia, 2000). Consequently, besides the most abundant variety, termed as N-MORB or massive eclogite, kyanite-rich deformed and ferrotitaniferous eclogites can be further distinguished (Gil Ibarguchi et al., 1990;Mendia, 2000).

| PETROLOGY OF THE STUDIED HP ROCKS AND SAMPLING
The eclogites are medium to coarse-grained rocks, characterized by a slight-moderate foliation. They are mainly composed of garnet, omphacite, quartz, zoisite, rutile and zircon, whereas secondary diopsidic pyroxene and amphibole (hornblende) form symplectitic intergrowths with plagioclase as a replacement of omphacite and garnet (Gil Ibarguchi et al., 1990;Peucat et al., 1990;Vogel, 1967). Garnet forms euhedral to subhedral millimeter-to centimeter-sized crystals with abundant mineral inclusions (quartz, rutile, zoisite and amphibole) in the core and inner zones (Gil Ibarguchi et al., 1990). Needle-shaped rutile inclusions in garnet were considered as outlines of a former igneous phase (clinopyroxene or titanite) by Gil  and Vogel (1967). Garnet porphyroblasts are characterized by almandine-pyrope-grossular solid solutions (Alm 30À52 Prp 25À40 Sps 0À2 Grs 20-25 ,) and display a weak chemical zoning with a decrease of Ca together with a slight Fe and Mg increase from core to rims, indicating prograde zonation (Gil Ibarguchi et al., 1990;Mendia, 2000).
During systematic sampling, we collected 38 samples (18 eclogites and 20 granulites) from 17 outcrops ( Figure 1). This study focuses on rock samples from three outcrops (one eclogite, one ultramafic and one intermediate granulite) being the most suitable for fluid inclusion study ( Figure 1). Additional 18 samples of pyroxenites from 6 outcrops have been collected from both Herbeira and Limo Massifs of the COC.

| ANALYTICAL METHODS
From the selected eclogite, granulite and pyroxenite samples from the COC, 60-100 μm thick and 30 μm doubly polished thin sections have been prepared. Fluid inclusion petrography was carried out in the Lithosphere Fluid Research Lab, Eötvös University, Budapest, using a Nikon OptiPhot2 optical microscope equipped with a Nikon CoolPix DS-Fi1 camera. The details on the analytical methods concerning SEM-EDS, Raman spectroscopy and FIB-SEM analyses are reported in supporting information (Appendix S1). The applied procedure to calculate the bulk MFI composition is provided in detail in Appendix S2 and Tables S3-S10. Mineral abbreviations are after Whitney and Evans (2010), except for specific and fictive endmembers of solid solutions (in Tables 3, S12 and S14) from Holland and Powell (2011) database and related works as used by Perple_X.
Phase equilibria and electrolyte fluid speciation, using bulk compositions of MFI from eclogite and ultramafic granulite, were computed by Gibbs energy minimization using the PerpleX algorithm (Connolly, 1990(Connolly, , 2009Connolly & Galvez, 2018). The endmember thermodynamic dataset and activity-composition models used are given in Appendix S1. Thermodynamic modelling focuses on the MFI bulk compositions of the eclogite and ultramafic granulite as having the most robust data from these samples (Table 2). Pseudosections were computed assuming a molecular fluid model, whereas electrolytic model, using the lagged minimization algorithm, was calculated at the two representative HP-HT and low-pressure-lowtemperature (LP-LT) conditions for comparison ( Figures 10 and S5 and Tables 3 and S12) and when estimating the potential original fluid in equilibrium with the bulk eclogite composition ( Figure S1 and Tables S14 and S15). Forward thermodynamic modelling has been performed in the CaFMAS-COHN system using elements as thermodynamic system components (Ca-Fe-Mg-Al-Si-C-O 2 -H 2 -N 2 ) to investigate solid phase relations with a COHN fluid. Fluid saturation conditions were not imposed during the computation; for further details see Appendix S1. The potential original composition of the fluid in equilibrium with the eclogite-using bulk rock data from by Gil , sample 97-was estimated using both molecular and electrolyte fluid speciation (details in Appendix S3).

| Petrography
The studied eclogite (COC17-062) is coarse-grained, moderately sheared and mostly made up by garnet and omphacite ( Figure 2a). In minor amounts, quartz, zoisite and rutile together with retrograde amphibole, chlorite and plagioclase-diopsidic pyroxene symplectite are also present. Garnet porphyroblasts are found as euhedral to subhedral 0.5-10 mm crystals, whereas omphacite with a size of 100-800 μm is usually anhedral, slightly elongated and commonly characterized by plagioclase-diopsidic pyroxene symplectite on the edges. Garnet crystals exhibit a characteristic zoning with respect to color and occurrence of inclusions. The following crystal zones were identified as follows: (1) core (zone 1) shows a dark pink color with abundant oriented needle-shape rutile, a few zircon and apatite inclusions, and (2) mantle (zone 2) exhibits light pink color and can be identified as a zone enriched in quartz and rutile inclusions, and the rim of garnet (zone 3) is nearly colorless and contains very few or no crystal inclusions (Figure 3a). Primary fluid inclusions occur within the growth zone 2 of garnet ( Figure 3b).
Among the various granulite subtypes in the COC, ultramafic and intermediate granulites have been selected for this study as containing fluid inclusions in significant quantities. The studied ultramafic granulite (COC17-079a) is a slightly sheared granoblastic rock, which is mostly made up by euhedral garnet (up to 1 mm in diameter) and anhedral clinopyroxene, without plagioclase ( Figure 2b). The intermediate granulite (COC17-060b) is foliated and is strongly retrogressed, being dominated by plagioclase, garnet and partially to completely amphibolized clinopyroxene (Figure 2c). Subhedral garnet crystals (0.1-1 mm in size) have a shape-preferred orientation forming porphyronematoblastic texture with plagioclase, clinopyroxene, amphibole, zoisite, quartz, rutile, zircon and opaque minerals. The retrograde assemblage in both types consists of hornblende AE actinolite, biotite, chlorite, zoisite and plagioclase-clinopyroxene symplectite. Garnet porphyroblasts in both granulite types have several cracks and have subhedral to anhedral grain boundaries indicating breakdown reactions. Despite this, core and inner zones are commonly undisturbed and contain primary MFI commonly together with crystal inclusions such as quartz, rutile and zircon.
The selected pyroxenite sample (COC17-026d) has a medium to coarse grained porphyroclastic texture with a slight foliation. It mostly consists of clinopyroxene with minor amounts of amphibole, orthopyroxene, olivine, spinel and serpentine ( Figure 2d). Clinopyroxene porphyroclasts, commonly partially replaced by amphibole along cleavages, are 0.5-5 mm in size, whereas serpentine associated with veins are widespread and overprint the development of each rock-forming mineral ( Figure 2d). Clinopyroxene, showing a lower degree of serpentinization, hosts secondary fluid inclusions. Graphite has not been identified in the rock matrix of any of the rock types.  Figure 3c). Rutile, quartz and zircon inclusions are commonly associated with these 3D clusters. MFI in granulites are smaller in size than those in eclogite (3-20 μm) and have negative crystal shapes (Figure 3d), whereas bigger inclusions are partially or fully decrepitated indicated by spherical-irregular decrepitation halos and comprising only solid phases.
Petrography and SEM-EDS analyses of exposed inclusions showed that MFI in both rock types are built up by aggregates of multiple micron-submicron-sized solid phases (Figure 3f,g); some of them (i.e. carbonates) are characterized by high birefringence under cross-polarized light. Carbonates, phyllosilicates, quartz and corundum have been detected in exposed MFI (Figure 3f,g). Carbonates consist of several minerals based on the variation in the ratio of Fe, Ca and Mg within a single multiphase fluid inclusion in each rock type (Table S2). Phyllosilicates including chlorite (Fe-Al-silicate) have been observed exclusively in MFI from eclogite, whereas pyrophyllite (Al-silicate) occurs in eclogite and both type of granulites (Table 2). Quartz and corundum are also present in MFI of ultramafic granulite.
In the investigated pyroxenite, secondary inclusions occur along healed fractures of clinopyroxene. The trails of fluid inclusions crosscut the orientation of amphibole lamellae in clinoproxene, typically running out to grain boundaries. The inclusions are smaller than 5 μm and show irregular and elongated shape ( Figure 3e). Optical microscopy revealed the presence of at least one tiny (up to 1-2 μm) solid mineral indicated by high birefringence.

| Raman spectroscopy
Raman spectroscopy revealed that MFI in eclogite and granulites show similar but distinct assemblages in terms of solids in the MFI. The solid phase assemblage in MFI in eclogite is Fe- These minerals show highly similar characteristic in terms of Raman spectra; therefore, their properties are discussed together with indications of differences as well. Representative Raman spectra of each phase are provided in Figure 4. Carbonates and phyllosilicates, dominated by pyrophyllite (262,195,705,216,113,814 and 3,671 cm À1 ) are always present; however, Fe-chlorite occurs only in MFI from the investigated eclogite, whereas margarite occurs in the MFI of investigated intermediate granulite. Identification of Raman bands of phyllosilicates, like pyrophyllite, Ferich chlorite and margarite was carried out using the data from Wang et al. (2015). Margarite is present exclusively in MFI from intermediate granulite with a strong spatial overlap with pyrophyllite or in small amounts within it. The characteristic feature of margarite is a low intensity band appearing around 393 cm À1 , together with an additional band at 3,633 cm À1 in addition to the OH À vibration mode of pyrophyllite (3,671 cm À1 ) ( Figure 4). Chlorite (main Raman bands at 336, 471, 554 and 661 and a triplet OH peak at 3,450, 357 and 3,625 cm À1 ) is associated with pyrophyllite. Raman spectra of carbonates ( Figure 4) indicate the presence of at least two types of Ca-Fe-Mg-carbonates with non-end-member mineral chemistry (Gillet et al., 1993;Rutt & Nicola, 1974). Calibrations for defining compositions from Raman spectroscopy proposed by Boulard et al. (2012) and Rividi et al. (2010) indicate a continuous shift of Raman bands between the two end-members of solid solutions (calcitemagnesite-siderite-rhodochrosite and dolomite-ankeritekutnahorite series). Therefore, as a first approximation, they can be considered as siderite (1,087, 292, 187, 730 and 1,732 cm À1 ) and ankerite (1,093, 283, 168, 722, 1,741 and 1,438 cm À1 ) as the best fitting mineral species. In small inclusions (below 8-10 μm), carbonates are difficult to distinguish from each other, because they are finely intergrown, based on the nearly continuous transition of characteristic Raman bands. Even for larger inclusions, their spatial distribution shows a broad overlap. To make Raman image analyses and further calculations unambiguous, carbonates will be considered hereinafter together.
Corundum and quartz occur together in MFI of ultramafic granulite. Corundum was also confirmed in MFI of eclogite in a few cases. Apatite, quartz and rutile occur occassionally in MFI and commonly as individual crystal inclusions in the host garnet. Raman bands with low intensities at 340, 372 and 424 cm À1 indicate the presence of pyrite observed as small spots at the wall of some of the inclusions. Graphite, with intense Raman bands at 1,341 and 1,605 cm À1 , occurs as few micron-submicronsized spots distributed together with the fluid phase.
Solid phases in secondary MFI in pyroxenites have also been identified as carbonate (calcite: 1,086, 281, 158 and 712 cm À1 ) and phyllosilicate (talc: 190, 114, 3,675 and 3,661 cm À1 ) ( Figure S2). In addition, these inclusions are volumetrically dominated by fluid phases such as nitrogen and to a lesser extent methane, so they can be also regarded as MFI.
Approximate modal proportions of dominant solid and fluid phases are calculated based on pixel counts of each phase on Raman images (Figure 6), indicated in Table 2 serial sectioning on selected representative MFI from each eclogite and granulite sample (Table 2).

| Gradual exposure of MFI with FIB-SEM
Considering data of MFI obtained with Raman imaging, FIB-SEM analyes were carried out on four representative MFI (two from eclogite, one from intermediate and one from ultramafic granulite). In accordance with Raman imaging on the same MFI, carbonates, phyllosilicates, corundum, quartz, graphite and pyrite were observed in tight intergrowths (Figure 7a,c,d). Each of them shows direct contact to the host garnet. Chemical variability in Ca, Fe and Mg of euhedral carbonates is visible on backscattered electron images (Figure 7a,c,d). Two different types of carbonates can be distinguished on BSE images: (i) Ca-rich Fe-Mg-bearing ankerite and (ii) Fe-rich Ca-Mg-bearing for siderite, with an estimated ratio of similar as it was inferred based on Raman spectra ( Figure 4). Although, to make further calculations straightforward, carbonates are treated together with representative compositions with respect to their estimated relative volume T A B L E 1 Relative proportions (in mol%) of fluid phases determined by Raman spectroscopy inside unexposed MFI in the studied rocks The complete dataset is presented in Figure 5. 'aver' refers to the average value for fluid composition in each rock type. Abbreviation: max, maximum; MFI, multiphase fluid inclusion; min, minimum; n, number of analyses.
F I G U R E 5 Distribution diagrams showing the relative proportion of components in the fluid phase (in mol%). Abbreviation 'av.' refers to the average of measurements in each rock type ratios in MFI, that is, 56:44 for ankerite and siderite, respectively (in MFI in ultramafic granulite and 48:52 for ankerite and siderite, respectively). The morphology of pyrophyllite is fibrous and lamellar, whereas in eclogite, it forms an intergrowth with chlorite (Figure 7a-c).
In intermediate granulite, pyrophyllite is interlayered with a Ca-bearing phyllosilicate (Figure 7f), confirming the occurrence of margarite in agreement with Raman spectroscopy (Figure 4). Characteristics SEM-EDS spectra of the dominant mineral phases are provided in Figure S3. On the basis of SEM imaging, MFI have numerous small cavities between solid phases and at the wall of the inclusion (Figure 7d), which corresponds to the location of the fluid phases ( Figure 6). Approximately 10-50 nm sized spherical-hexagonal particles of graphite have been found in some of these cavities ( Figure 7d). Furthermore, pyrite with traces of Cu and As has been observed at the walls of the inclusions with a size ranging between 100 and 300 nm. A few submicron-sized (100-200 nm) spots of apatite and rutile (accidentally trapped minerals) have also been detected. Application of SEM image analyses with labeling of distinct phases on each equidistant slices allowed 3D reconstruction of the inclusions and determination of  Table 2). Volume percentages in eclogite (COC17-062, Figure 8a,c) are as follows: 29%-31% carbonates, 30%-34% pyrophyllite, 17%-18% chlorite and 18%-23% cavity (fluid), whereas graphite, sulfide and accidentally trapped solid phases do not exceed 1%. In the ultramafic granulite sample (COC17-079a, Figure 8b,e), the following volume distribution has been acquired: 34% carbonates, 6% pyrophyllite, 8% corundum, 21% quartz and 31% cavity (fluid), whereas the volume of pyrite and other accidentally trapped minerals is <0.5%. MFI in the intermediate granulite sample (COC17-060b, Figure 8c) is composed of 30% carbonates, 33% pyrophyllite, 2.5% margarite and 34% cavity (fluid), whereas pyrite and accidentally trapped solid phases do not exceed 0.5%. Corundum was also confirmed with Raman spectroscopy in some MFI from eclogite, but it has not been identified during serial sectioning, except in the ultramafic granulite sample (Figure 7e).

| Approach used to estimate bulk MFI composition
We estimated the bulk compositions of individual MFI in terms of the major elements using the modal proportions of the solid and fluid phases in the MFI and their composition obtained by means of EDS-SEM. Modal proportions of the phases in the MFI were obtained by Raman 2D imaging (n = 30 and n = 9, for the eclogite and ultramafic granulite hosted MFI, respectively) under the assumption that relative areas can be extrapolated to volume proportions. This assumption is supported by the good agreement between area and volume proportions obtained by either Raman imaging or FIB sectioning (see Tables 2 and S3) on selected MFI. In detail, the procedure we applied was as follows: (1) computing the mass of an (arbitrary) 100 cm 3 of all investigated MFI (Table S4, column "Mass g/100 cm 3 ") using the normalized vol.% (taken from the area %) of the identified phases (Table S4) and the density of the solids at ambient conditions (taken from the Holland & Powell, 2011 database, Table S8) and the density of the residual fluid phase (Table S7) assumed to be at a pressure of 500 bar based on Raman shifts (Tables S10 and S11, see also more details in Appendix S2), (2) computing the volume per kg of MFI of individual phases and the fluid (Table S5) by normalizing the vol.% of the individual phases to one kg using the total mass for 100 cm 3 of MFI obtained in the previous step, (3) expressing the amount of individual phases in mol/kg of MFI (Table S6) using the molar volume for individual phases and endmembers of the gas at 500 bar (cm 3 /mol, Table S8) and finally, (4) the mol/kg of the components Ca-Mg-Fe-Al-Si-H 2 -O 2 -C-N 2 for the MFI (Table S9) was found by using the atoms per formula unit of the mineral phases and the fluid molar composition T A B L E 2 Estimated relative proportions of solid and fluid phases based on Raman imaging and FIB-SEM analyses on primary MFI in garnet of eclogite and granulites from COC   ) showing an intergrowth of carbonates and phyllosilicates (margarite is interlayered with pyrophyllite). Dark areas are voids exposed during serial sectioning, which contain a fluid phase (N 2 AE CH 4 AE CO 2 ), presented in Table 1 and Figures 4 and 5 (Table S8). Details about the choice of molar volume for solids and the residual fluid can be found in the supporting information (Appendix S2). The use of elements instead of conventional weight of oxides to define the bulk composition is preferred when using electrolytic thermodynamic modelling (Connolly & Galvez, 2018). The pseudosection presented in the main text ( Figure 10) was computed assuming a simple molecular fluid model, but for comparison, we present the results of the electrolytic model using the lagged minimization algorithm for two representative HP-HT and LP-LT conditions (Table 3).

| Origin of MFI and fluid composition
Room temperature phase assemblages of the studied MFI show similar features of multiphase solid inclusions (MSI) described in UHP rocks (e.g. Frezzotti & Ferrando, 2015;Philippot & Selverstone, 1991). MSI are thought to represent a homogeneous solute-rich supercritical aqueous fluid at entrapment (Ferrando et al., 2005;Frezzotti & Ferrando, 2007;Maffeis et al., 2021;Hermann et al., 2006;Zheng & Hermann, 2014). By contrast, MFI from this study comprise a ubiquitous and constant presence of fluids in the studied rock types (Tables 1 and 2). Additionally, MSI have been recorded only in UHP rocks (Frezzotti & Ferrando, 2015), whereas peak metamorphic conditions obtained in eclogites from the COC did not exceed 2.2 GPa (Mendia, 2000) and 1.7 GPa for granulites (Puelles et al., 2005), thus making less likely the direct precipitation from oversaturated aqueous fluid (daughter origin) as the only mechanism for producing the observed MFI. Multiphase inclusions were previously recorded in HT granulite facies rocks that experienced anatexis (Carvalho et   are defined as crystallized silicate melt, thus indicating fluid-melt immiscibility (Cesare et al., 2015;. In our study, however, no petrographic evidence of partial melting and no nanogranitoid inclusions were observed. Thus, it is unlikely that MFI characterized in this study are products of fluid-melt immiscibility.
Textural observations strongly suggest that MFI in garnet of eclogite and granulites from the COC have primary origin (Figure 3) and indicate entrapment during garnet growth likely at prograde-to-peak metamorphic stage (Frezzotti & Ferrando, 2015;Roedder, 1984). Similar but distinct solid phase assemblages of MFI (dominated by carbonates and phyllosilicates, Figure 4) and their nearly constant volume proportions (Table 2) indicate that a homogeneous fluid was trapped in the garnets. This is further supported by the fact that neither carbonates nor phyllosilicates have been found in the matrix rock as rock-forming minerals or crystal inclusions independently of MFI. Furthermore, solid phases in MFI have the same major to minor elements (Fe, Mg, Ca, Al and Mn) as that of the host garnet (Table S2). All of these suggest that minerals in the studied MFI are the result of a post-entrapment reaction of the trapped fluid with the host garnet (step-daughter minerals). The possibility of different post-entrapment modifications, local fluid-host mineral interactions and re-equilibration producing stepdaughter crystals during exhumation processes has already been proposed for fluid inclusions in HP metamorphic rocks (Bodnar, 2003;Frezzotti & Ferrando, 2007;Maffeis et al., 2021;Philippot & Selverstone, 1991;Roedder, 1984;Scambelluri & Philippot, 2001;Touret, 1981Touret, , 2001. Chemographic analysis of the MFI bulk chemical composition in eclogite and ultramafic granulite ( Figure 9 and Table S9) shows that they are compatible with a garnet + COHN fluid composition (i.e. a mix between garnet and the line joining CO 2 -H 2 O on Figure 9b). Some of the bulk MFI compositions containing significant enrichment of graphite (5%-12% in modal proportions based on Raman imaging) fall out of the projected garnet-CO 2 tieline (Figure 9). Graphite content may be significantly overestimated during measurements with Raman spectroscopy, due to its fairly high Raman activity, in agreement with very low bulk graphite contents obtained by means of FIB analyses in MFI from eclogite. Modal proportion of graphite based on Raman imaging was estimated to be around 10 area % in two MFI from eclogite, whereas FIB analyses revealed that the real graphite content of the same inclusions is below <1 vol.% ( Table 2). Overestimation of graphite can be explained by its high Relative Raman Cross Section, as a similar phenomenon of different gases and liquids (e.g. Haskin et al., 1997;Schrötter & Klöckner, 1979). Therefore, the graphite-enriched MFI compositions are considered biased by the Raman measurements and are not further considered.
The observed assemblage of the MFI in eclogite can be reproduced with comparable volume proportions at 0.5 GPa and 320 C ( Figure 10) by thermodynamic modelling (molecular and electrolytic model) using the bulk composition determined by the FIB mineral proportions (MFI_FIB1 on Figure 9 and Table S9) as being a representative example for the MFI (Figure 10a,b and Table 3). Moreover, the modelled carbonate compositions in this low-pressure assemblage have similar compositions to the ones observed in the MFI based on EDS data (ank 58 dol 42 vs. ank 50 dol 50 and sid 80 mag 20 vs. sid 70 mag 30 for modelled and observed, respectively, Table 3). The model predicts a high amount of quartz, which is, however, in disagreement with most of the observations (Table 2). This discrepancy can be explained by the modelled chlorite composition, which is poorer in silica (2.0 Si apfu) relative to the measured composition (2.7-2.9 Si apfu; see Tables S2 and S8). The production of carbonates and hydrous silicates through the carbonation and hydration reaction of the almandine endmember in garnet with CO 2 and H 2 O was previously proposed by thermodynamic calculations in simple systems (Carvalho et al., 2020;Marini, 2007). At peak conditions (1.8 GPa, 800 C), the modelled assemblage is largely (>85 vol.%) formed by garnet and a COHN-fluid ( Figure 10b) supporting a homogeneous entrapment of a fluid composed mostly of H 2 O (56 mol%) and CO 2 (34 mol%) with additional N 2 (10 mol%, Figure 10c, Table 3). The occurrence of other minor minerals, such as clinopyroxene (5.2 vol.%), carbonates (6.0 vol.%) and graphite (2.6 vol.%) suggests either (1) a slightly incorrect bulk composition estimate from the FIB-based MFI data or (2) that the original fluid was more enriched in H 2 O (somewhere along the tieline joining the MFI and H 2 O, Figure 9), which can be explained by a partial H 2 O loss. Computations using either molecular fluid or electrolytic aqueous solutions do not significantly modify these volume proportions (compare Tables 3 and S12). Preservation of the original amount of H 2 O in inclusions formed at high P-T conditions is quite rare, together with the general process of selective H 2 O loss from fluid inclusions during exhumation (e.g. Frezzotti et al., 2012;Frezzotti & Ferrando, 2015;Hall & Sterner, 1993;Hidas et al., 2010;Maffeis et al., 2021). Various mechanisms of reequilibration (decrepitation, leakage by nanocracks/ dislocations or diffusion) during post-entrapment evolution of fluid inclusions may simultaneously operate and cannot be observed under optical microscopy (Bakker & Jansen, 1994;Bodnar, 2003;Roedder, 1984;Viti & Frezzotti, 2001;Yamamoto et al., 2011). Different degrees of leakage and decrepitation further depend on the size of the inclusions, that is, smaller inclusions are more prone to 'survive' without decrepitation (e.g. Campione et al., 2015). Loss of H 2 O may take place not only by optically visible or nanoscale decrepitation of inclusions but also by other processes such as passive diffusion to the host mineral due to the small size of hydrogen protons (Bodnar, 2003;Mavrogenes & Bodnar, 1994).
In addition to MFI in eclogite, thermodynamic modelling, using bulk composition of a representative multiphase fluid inclusion in ultramafic granulite, presented in Figure S4 (MFI_FIB3 on Figure 9) confirms that post-entrapment reactions are likely to have evolved in a similar way to that observed for the MFI in the eclogite.
The assemblage pyrophyllite + two carbonates (dolomite-ankerite and magnesite-siderite solid solutions) +quartz + kyanite + minor hematite (and residual N 2enriched fluid) can be reproduced at low P-T conditions similar to the one inferred for MFI in eclogite (Figure 10a). The absence of chlorite in the modelled assemblage also agrees with the observations. However, corundum instead of kyanite occurs in all studied MFI. The production of kyanite instead of corundum + quartz in the MFI may be controlled by the metastable behaviour of the aluminosilicate polymorphs due to the extremely small size of inclusions as discussed by Carvalho et al. (2020). At peak conditions, the modelled assemblage has dominantly a garnet and COHN-fluid composition (52.97 + 38.19 vol.%, respectively), together F I G U R E 9 Compatibility diagrams (chemographies) constructed after reducing the compositional space to four (a) and three (b) components (projection phases and exchange vectors are indicated), showing individual estimates of the bulk multiphase fluid inclusions (MFI) compositions for eclogite and ultramafic granulite (see Legend) based on molar proportions in the CaFMAS-COHN system with the measured compositions of carbonate, chlorite and pyrophyllite by EDS (for dataset see Table S2). Most of the values for MFI in eclogite fall into the chlorite-pyrophyllite-siderite-magnesite (CO 2 ) triangle lying between the endmembers of garnet and H 2 O-CO 2 fluid composition. All of those points falling out of the chlorite-pyrophyllite-siderite-magnesite (CO 2 ) triangle are MFI with high (5%-12% in modal proportions) and potentially overestimated graphite content ( Table 2). The lines ending with black squares are tielines joining the garnet and MFI with the graphite that due to the projection from methane it plots at the infinity. Dashed grey line indicates a trend of potential H 2 O loss, which is paralell with a line connecting bulk MFI data points with modelled fluid compositions in equilibrium with the eclogite. The purple circles correspond to these modelled fluid compositions using different CO 2 contents in equilibrium with the eclogite; for details, see text and Appendix S3 and Figure S1. Compatibility diagrams were plotted using CSpace (Torres-Rold an & Garcia-Sanchez, 2000) F I G U R E 1 0 (a) Pseudosection of a representative multiphase fluid inclusions (MFI) bulk composition in eclogite (MFI_FIB1 on Figure 9a,b). The stability fields of carbonates, chlorite, pyrophyllite and garnet are indicated (with colored lines). Color blue levels indicate the variance of the assemblage (the darker the higher the variance). Carb designates the presence/abscence of any of the two carbonates. Dashed grey line represents an iso-density path for the whole system (2,637 kg/m 3 ) along which the proportions of the solid and fluid phases are presented in Figure 10b,c. Retrograde PT-path of the rock and dashed rectangle indicate the range of D4 deformation phase recorded in the COC, after Puelles et al. (2005). (b) Distribution of volume proportions of the solid phases along the iso-density path for the whole system (2,637 kg/m 3 Figure 10a) of MFI_FIB1 bulk composition in eclogite. Minor phases are consisted of kyanite, talc and chloritoide. Dashed grey lines on (b) and (c) indicate the beginning of fluid immiscibility along the retrograde path, which may correspond with the stability zone of pyrophyllite. Modal proportions, mineral composition and phase speciation of stable phases at two representatives highpressure-high-temperature (HP-HT) and low-pressure-low-temperature (LP-LT) conditions are presented in Tables 3 and S13. (c) Distribution of molar proportions of the fluid phases along the iso-density path for the whole system (2,637 kg/m 3 Figure 10a) of MFI_FIB1 bulk composition in eclogite. The fluid is in molar basis and is the average of the two immiscible fluids for the last part of the decompression path. Even though the enrichment of N 2 and consumption of CO 2 with decreasing temperature fits well with our observations in the MFI, the modelled H 2 O and CH 4 contents for low temperatures show a disagreement; furthermore, the change of CH 4 content could not be represented as having insignificantly low amounts in the model. This could be explained by the H 2 O loss, as well as our little constrain on oxygen fugacity during the retrograde path (for details see text) with other minor phases such as dolomite (4.72 vol.%), kyanite (1.66 vol.%) and quartz (2.46 vol.%), which indicate a similar homogeneous entrapment together with an even higher degree of partial loss of H 2 O.
The formation and development of secondary MFI in pyroxenites may be possibly explained by similar postentrapment processes affecting an originally trapped H 2 O and CO 2 -bearing fluid. The host mineral, nearly pure diopside (Tilhac et al., 2016, could have allowed step-daughter phases to take-up elements like Ca and Mg, thus production of calcite (CaCO 3 ) and talc (Mg 3 Si 4 O 10 [OH] 2 ) via carbonation and hydration reactions, possibly as analogous mechanism evidenced in orthopyroxene described by Frezzotti et al. (2012).
To have an independent constraint on the potential original composition of the trapped fluid, we made a comparison to assess if the obtained fluid composition in eclogite (H 2 O: 56 mol%, CO 2 : 34 mol% and N 2 : 10 mol%) at HP from the MFI bulk composition hosted in garnet could be compatible with a fluid in equilibrium with the bulk eclogite assemblage (Appendix S3 and Figure S1). This was calculated based on bulk rock data of Gil  and implies that a value higher than 25 mol% CO 2 content in the fluid is compatible with the eclogite mineral assemblage devoid of zoisite (plotted on Figure 9) at peak conditions (i.e. 800 C and 18 kbar; see Appendix S3, Figure S1 and Table S14-15 for futher details). The results of thermodynamic modelling on the composition of the trapped fluid at HP-HT conditions (Figure 10c and Table 3) suggest a CO 2 content of 34 mol %, indicating a CO 2 -bearing aqueous character. This fluid is characterized by a certain solute charge (2.8 mol/kg, Table S15) and could have been indeed in equilibrium with the bulk eclogite. Thus, there is no requirement for influx of an external fluid, or alternatively, the external fluid, if any, should have equilibrated with the bulk eclogite before the entrapment.

| Formation of solids within the MFI
As stated above, carbonates and sheet silicates within the MFI (Figures 3, 4, 6 and 7) are clearly step-daughter minerals. Nevertheless, minerals (i.e. quartz, rutile and apatite) that are both present in MFI and in the host garnet as crystal inclusions (Figure 3b) can be considered partially as accidently entrapped minerals in each rock type. However, the presence of quartz with corundum ( Figure 7e) was confirmed in the majority of MFI in ultramafic granulite. This indicates that in this rock, quartz can also be interpreted as a step-daughter mineral, similarly as described by Carvalho et al. (2020) and Tacchetto et al. (2019). According to the prediction of our model (Figure 10b and Table 3), such origin of quartz is possible in the MFI in eclogite too, without considering accidental entrapment. Furthermore, solid phases in the MFI could have also precipitated from the COHN fluid in minor fractions. Enhanced solubility of elements, like Si, Al or Fe, in our case, in H 2 O at HT-HP conditions has been proposed by several authors (e.g. Dolejš & Manning, 2010;Kessel et al., 2004;Manning, 1994;Newton & Manning, 2003) accounting for the possible crystallization of daughter minerals directly from the trapped fluid. It may be especially true for quartz due to high solubility of Si in H 2 O at these conditions. This is also supported by the non-trivial charge of solute computed for the eclogitic assemblage at peak conditions (approximately 2.8 mol/ kg, Table S15). Interestingly, the modelled solute charge of the new system (after entrapment) C-bearing fluidgarnet is significantly lower (approximately 0.184 mol/kg at peak conditions and nearly 0 at 320 C and 5 kbar, Table S13). It is highly likely that the trapped fluid, equilibrated with the eclogite, would additionally attain equilibrium with the host garnet. This process should result in precipitation of some of the observed solid phases, like quartz within MFI, as being one of the most abundant aqueous species in the trapped fluid (Table S15). However, it can only represent a minor fraction of the observed MFI assemblage as the peak conditions of the COC eclogite (18 kbar and 800 C, Gil Ibarguchi et al., 1990) suggest maximum solubilities of approximately 6 wt.% of Si in the aqueous fluid (Tables S14 and S15), whereas an overall 20 wt.% of C-bearing fluid supposedly reacts with 65 wt.% of garnet (Table 3). Our calculations nevertheless clearly indicate that bulk chemical composition of the studied MFI ( Figure 9 and Table S9) is consistent with a garnet + COHN fluid composition. This implies that step-daughter producing postentrapment reactions can be considered as plausible dominant processes, which are responsible for the formation of the MFI, without excluding the possibility of minor production of daughter minerals.

| Timing of MFI crystallization
According to the observed intricate intergrowth of the solid phases (Figure 7), it is highly likely that the mineral assemblage in the MFI precipitated together in a single step as there is not any sign of systematic zonation or a traceable sequence of solid phases inside the MFI (Figure 7a,c), which would be a result of sequential crystallization or periodic precipitation as evidenced in MSI in pyrope from Dora Maira whiteschist (Frezzotti & Ferrando, 2015). Under equilibrium conditions, carbonates and chlorite would be expected to crystallize earlier at temperatures higher than pyrophyllite (> 600 C,Figure 10) in MFI in the eclogite. This seems unlikely as (1) microstructure of the inclusions does not support this interpretation (see above and Figure 7) and (2) there is no trace of other solid phases in the observed assemblage, such as kyanite, talc or chloritoide which all would appear at higher temperatures if equilibrium would have been attained continuously during decompression ( Figure 10). Thus, it is inferred that carbonates and chlorite have possibly remained at metastable state until reaching the stability field of pyrophyllite. Graphite is predicted to be stable in a low-pressure assemblage together with carbonates, pyrophyllite, chlorite and quartz; even the MFI bulk composition used for modelling was computed from a graphite-free assemblage. This type of metastable behaviour with delayed nucleation and crystallization may be possibly due to the small volume of the inclusions as suggested by Carvalho et al. (2020), together with other effects modifying the energy of the system, such as surface tension rather than the bulk Gibbs free energy. It cannot be ruled out that the metastable crystallization could be linked to or even triggered by the H 2 O loss from the MFI and/or a deformation episode during exhumation. This latter possibility may correspond with a deformation phase, termed as D4 phase with extensional shear byÁbalos et al. (1996) and Puelles et al. (2005) during late retrogression of the HP units of the COC (Figure 10a). A fast reequilibration from a metastable state concomitant with the partial loss of H 2 O could further explain the occurrence of CO 2 in the residual fluid that should have been consumed during carbonation at higher temperature (min. 500-700 C) than the maximum temperature stability of pyrophyllite (around 400 C) ( Figure 10).
A further notable prediction of the thermodynamic model is that at temperatures below 350-400 Ccorresponding to the stability field of pyrophyllite ( Figure 10)-the residual fluid separates into two immiscible fluids: one fluid (F1) is nearly H 2 O pure (approximately 94 mol%) and the other (F2) is a CO 2bearing and is highly enriched in N 2 (up to 60 mol%) (Figure 10c and Table 3). This temperature range corresponds with the extended immiscibility field of the H 2 O-CO 2 system for a wide P-TX range (Diamond, 2001;Hurai, 2010). An immiscible H 2 O-rich fluid phase in the inclusion would further enhance both H 2 O-loss from the MFI and sheet silicate producing reactions due to the increased activity of water. Wetting properties of H 2 O account for distribution along the wall of the inclusion, easing its escape as a single phase through nanocracks/decrepitation. Equilibrium decompression path should produce a fluid, which is progressively enriched in H 2 O (Figure 10c) contrary to the observations. When entering the stability field of pyrophyllite, there is an abrupt change in the composition of the fluid with CO 2 decrease, together with an increase of N 2 . Nitrogen content is passively increased in the fluid due to the crystallization of carbonates and phyllosilicates. Enrichment of CH 4 (up to 87 mol% in the residual fluid phase) was also observed in MFI from eclogite (Table 1); nevertheless, its production is highly sensitive to certain redox conditions in the inclusions and precipitation of graphite (see below).
The lack of detectable H 2 O as a fluid phase in the MFI, both in the eclogite and the granulites, could indicate that the fluid-garnet interaction ceased when H 2 O was consumed in the reaction and hydration reaction produced pyrophyllite. Thermodynamic modelling (Figure 10c), however, suggests that CO 2 is almost fully consumed during the retrograde carbonation reactions; meanwhile, N 2 -H 2 O-enriched residual fluid is in equilibrium with the retrograde assemblage and is compatible with H 2 O-saturated conditions. Similarly, enrichment of nitrogen in the fluid phase is also observed in the secondary inclusions in the studied pyroxenites of the COC (Figure 3e).

| Fluid regime at HP subduction zone conditions
Fluid inclusions witnessing N 2 -rich retrograde and near peak fluids have been described in the Caledonides of Western Norway (Andersen et al., 1990(Andersen et al., , 1991Jamtveit et al., 1990). Isochoric calculations of Andersen et al. (1991) indicate that the earliest high-density N 2 -rich AE CO 2 fluid inclusions, trapped secondarily in quartz from eclogitic veins, represent early retrograde metamorphic fluids, possibly rather close to peak conditions (around 1 GPa and 600 C). Late-stage secondary fluid inclusions are rich in N 2 , which are probably the result of decrepitation and redistribution of early fluid inclusions (Andersen et al., 1991). Although the presence of nitrogen-bearing fluids is a general feature of HP metamorphism (Andersen et al., 1989(Andersen et al., , 1990(Andersen et al., , 1991Tomilenko & Chupin, 1983;Touret & Frezzotti, 2003), possible processes of abiotic N 2 enrichment in the fluid are not well constrained.
Fluid-rock interaction, similar to the reactions observed in the MFI, is highly likely to have affected the untrapped fluids as well in the subduction zone during their retrograde path. Similar partial or complete consumption of H 2 O and CO 2 during retrograde evolution would designate a horizon (between 300 C and 400 C, below 1.0 GPa, Figure 11) of a fluid regime dominated or at least elevated in nitrogen, even though the nitrogen in the originally trapped fluid was probably present in minor amount.
The above-described fluid-rock interaction results in abiotic enrichment of not only nitrogen in the fluid phase but also methane especially in MFI in eclogite (comprising the residual fluid in 21-87 mol%). Such enrichment is not reproduced by our thermodynamic modelling where methane concentrations remain very low (Figure 10c and Table 3). The disagreement could be explained by our little constrain on oxygen fugacity during the retrograde path and for the neglected role of sulphur on redox conditions (Maffeis et al., 2021). Equally, the observed difference on the residual fluid composition between MFI of eclogite and granulite samples (Table 1) could be possibly explained by a difference of the initial composition of the fluid, that is, having various original ratio of H 2 O-CO 2 (-CH 4 -N 2 ) and potentially different oxygen fugacity conditions in the different lithologies. Residual fluid composition of MFI in eclogite shows a slightly wider compositional range in CO 2 -CH 4 -N 2 content ( Figure 5), which could be possibly due to small variations of N 2 in the trapped fluid and/or partial loss of H 2 O after entrapment, resulting in a change in COH content and speciation in the fluid phase. Diffusion of elements (i.e. H) from the inclusions may result in a significant modification of the fluid composition and change the oxygen fugacity of the inclusion (Frezzotti & Ferrando, 2015). However, oxygen fugacity and activity of carbon during trapping of the COH fluid along with post-trapping reequilibration and modifications may be potentially a  (2019); contours of the mantle wedge (green for the lithospheric mantle and red-purple for HP tectonic mélange) during exhumation are drawn after Gerya (2019) dominant factor for causing variations on the ratio of residual CO 2 and CH 4 in the inclusions (Hall & Bodnar, 1990). Methane, despite being quite uncommon, together with N 2 and CO 2 has been also identified as a significant constituent of prograde and peak metamorphic fluids in eclogite and granulite facies rocks (Carvalho et al., 2019;Fu et al., 2003;Maffeis et al., 2021;Mukherjee & Sachan, 2009;Tao et al., 2018). Abiogenic production of methane corresponds to reduction of carbonates (as the most abundant carbon-bearing phase in subduction environment) in the slab at HP-HT conditions with water saturation (Tao et al., 2018). Furthermore, the speciation of a carbon saturated COH fluid is controlled by the certain redox conditions of the system (Connolly, 1995;Galvez et al., 2013Galvez et al., , 2016Hall & Bodnar, 1990).
The presence of graphite in some of the MFI indicates that the trapped COH fluid became graphite saturated locally leading to its precipitation from the residual fluid during cooling and exhumation (Carvalho et al., 2019;Cesare, 1995;Cesare et al., 2007). As graphite coexists with methane-bearing residual fluid which is always connected to the small cavities between solid phases of the inclusions (Figure 7d), we assume that the fluid reached a saturation after the carbonation and hydration reactions and locally became graphite saturated in some of these cavities. This could have resulted in precipitation of graphite in some of the parts filled with residual fluid, explaining also small variations of fluid composition measured in a single multiphase fluid inclusion.

| Implications for nitrogen cycling
The behaviour of nitrogen during surface-subsurface biochemical reactions is relatively well understood; however, the knowledge about its global planetary cycle and the role of subduction as well as storage need further constraints Cannaò et al., 2020;Cartigny & Marty, 2013). Most studies agree that nitrogen input into the deep Earth is greater than outgassing (Busigny et al., 2011;Mitchell et al., 2010;Mysen, 2019;Sano et al., 2001), although Fischer et al. (2002) suggested a comparable extent of nitrogen subducted in sediments (i.e. clay minerals) versus nitrogen degassed by arc volcanism at the Central America convergent margin. As previously discussed, nitrogen is a commonly encountered constituent of fluids in subducted eclogite facies rocks. In the geological cycle of nitrogen, the initial source introduced into the subduction zone is organic material deposited in seafloor sediments . During diagenesis and low-grade metamorphic evolution, nitrogen is released into pore fluids as ammonia (NH 3 ) then transforms into the form of ammonium (NH 4 + ) (Busigny & Bebout, 2013;Cedeño et al., 2019). Having similar charge and ionic radius, the released NH 4 + ion acts as an ideal substitution for K + in such phyllosilicates as clay minerals and micas, and alkali feldspars, confirmed by strong correlation between concentration of nitrogen and LIL elements, such as K, Rb and Cs (Busigny et al., 2003;Halama et al., 2010). The breakdown of micas during prograde metamorphism results in a continuous release of NH 4 + , although high stability of NH 4 + under HP in white micas (phengite) or K-feldspar (hollandite) allows its retention to great depths (even of the transition zone) depending on redox conditions (Bebout, 2007;Cedeño et al., 2019;Watenphul et al., 2009). Efficiency of nitrogen release, which would equilibrate as N 2 (Li & Keppler, 2014), is mostly controlled by the breakdown of phengite, which is showing a strong correlation with the thermal gradient of subduction zones (Mitchell et al., 2010). Consequently, hot subduction paths promote a high degree of N escape, whereas cold geotherms enable the retention of NH 4 + in HP micas or clay minerals resulting in efficient transport of nitrogen into great depths (Cedeño et al., 2019). Previous studies ofÁbalos et al. (2003) and Puelles et al. (2005) highlighted that the studied suites of HP subducted rocks of the COC have been involved in prograde metamorphic evolution following a warm subduction PT path with 13-14 C/km, which would be sufficient for elevated N 2 release (Busigny & Bebout, 2013). Bulk N 2 concentration and stable nitrogen isotopic composition of eclogites from the COC described by Halama et al. (2010) reflect no further metasomatic addition (K-Rb-Cs-Ba) during subduction; overlap of N 2 concentration and δ 15 N with data for the altered oceanic crust suites implies retention of seafloor-inherited N 2 to depth of around 90 km. Halama et al. (2010) further suggested that N 2 concentrations of 3-8 ppm hosted by fluid inclusions are required in COC eclogite because of the lack of phengite in these rocks. The present study confirms such inference (Figures 3b and 4). Accordingly, N 2 released from the slab would not fully return to the surface via arc volcanism, but HP subducted rocks can be a potential reservoir for N 2 -rich fluids at shallower environment of the exhumation (Figure 11).
Our findings indicate a front at the temperature of 300-400 C, which is characterized by the enrichment of nitrogen during the exhumation path of HP rocks. The release of N 2 -dominated fluids during retrograde processes may occur at shallower depths, resulting in enrichment of abiotic nitrogen in the fluid phase. This mechanism would serve to explain the presence of N 2 -rich fluids associated with HP serpentinization and would contribute to the elevated concentrations of nitrogen documented in serpentinized peridotites in subduction settings (Halama et al., 2012;Philippot et al., 2007;Vitale Brovarone et al., 2020). Prevalence of N 2 compared with H 2 O and CO 2 in the fluid phase is witnessed in secondary fluid inclusions of pyroxenites as well, in accordance with retrograde metamorphic fluid compositions in the Caledonides (Andersen et al., 1991). Released or remobilized N 2 -rich fluids may be channelized, transported and upflow providing contribution to the fluid regime of shallow subduction depths and even the subsurface environment (Figure 11). This conclusion is in good agreement with N 2 -rich fluid composition of gas outflow of mud volcanoes and thermal springs at forearc regions of accretionary subduction complexes (Giggenbach et al., 1993;Lee et al., 2017;Motyka et al., 1989;Snyder et al., 2003), with reaching even 92 vol.% reported, for example, at the Andaman and Nicobar Islands (Chaudhuri et al., 2012). All of these suggest that such significant addition of nitrogen with deep subduction origin should play an important role in global nitrogen cycling and may provide significant contribution to N 2 supply to the overlying subsurface/surface biosphere during devolatilization in the forearc regions of convergent margins.

| CONCLUSIONS
The investigated primary MFI in eclogite and granulites and secondary MFI in pyroxenites from the COC represent trapped COHN fluids during prograde-to-peak metamorphism that were significantly modified by reactions with the host garnet during exhumation. Our results record a detailed characterization of microstructural, microchemical features, origin and post-entrapment evolution of primary MFI, together with thermodynamic modelling. The main implications of our study are summarized as follows: • Trapped fluids can be described in the COHN (1) the formation of carbonates and phyllosilicates, (2) the passive enrichment of nitrogen in the residual fluid observed in the studied MFI (N 2 content: 13-68 mol% in the observed MFI) and (3) local enrichment of methane in MFI in eclogite (varying from 21 to 87 mol%). • Thermodynamic models revealed that step-daughter assemblage is stable at low-pressure and lowtemperature (between 2 and 9 kbar and 300-400 C) conditions. Our model and the microstructural data of the MFI suggest that the formation of the stepdaughter minerals is expected at the same time due to metastable behaviour at higher pressure and temperature.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of the article at the publisher's website. Appendix S1. Detailed description of the applied analytical methods: SEM-EDS, Raman spectroscopy, FIB-SEM and thermodynamic modelling. Appendix S2. Molar volumes used to estimate MFI bulk composition. Appendix S3. Calculation of a potential original composition of the fluid in equilibrium with the eclogite. Figure S1. A pseudosection and the iso-values for CO 2 in equilibrium with the eclogite to specify the effect of CO 2 addition on the mineral assemblages at peak conditions (dashed grey line), calculated based on bulk rock data of Gil , for details see text and Appendix S3. Modal proportions, mineral composition and phase speciation of stable phases are presented in Table S14-15. Figure S2. Representative Raman spectra of solid (calcite and talc) and fluid phases (nitrogen and methane) in MFI in clinopyroxene of pyroxenite from the COC. Figure S3. a-b-c) Representative EDS spectra of dominant solid phases of the MFI in the studied eclogite and granulites obtained during FIB sectioning. Figure S4. Pseudosection of a representative multiphase fluid inclusion in ultramafic granulite using the bulk composition of MFI_FIB3, shown on Figure 9. The stability fields of carbonates, pyrophyllite and garnet are indicated (with colored lines). Color blue levels indicate the variance of the assemblage (the darker the higher the variance). Carb designates the presence/abscence of any of the two carbonates. Table S1. Relative proportions (in mol%) of fluid phases determined by Raman spectroscopy inside unexposed MFI in the studied rocks, presented on Figure 5. Table S2. Mineral chemistry of the measured solid phases (carbonates, pyrophyllite and chlorite) in exposed MFI and the host mineral (garnet) of eclogite (COC17-062) and ultramafic granulite (COC17-079a) based SEM-EDS data. Table S3. Volume (vol.%) and modal (area%) proportion of solid and fluid phases (raw data) in MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Table S4. Modal (area%) proportion of solid and fluid phases (extrapolated to vol.% and normalized to 100%) in MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Density was calculated (see Table S7) according to estimation of residual pressure in the MFI using characteristic peak positions of the fluids, for details, see Appendix S2. Table S5. Volume per kg of system (cm 3 ) of solid and fluid phases in MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Table S6. Mol per kg of system (cm 3 ) of solid and fluid phases in MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Table S7. Fluid density calculated based on fluid mol% in MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Table S8. Compositional space and compositional corrected estimated densities. The densities of the fluid phases are computed at 500 bar and room temperature (see Appendix S2 for details). Table S9. Bulk composition of the system in mol/kg of MFI of eclogite (COC17-062) and ultramafic granulite (COC17-079a) obtained with FIB-SEM analyses (eclogite: MFI_FIB1-2 and granulite: MFI_FIB3) and Raman imaging (eclogite: MFI#1-30 and granulite: MFI#31-39), respectively. Table S10. Modal proportion of the fluids with Raman band positions in MFI in eclogite (COC17-062) applied for estimation the residual pressure in the inclusion. Table S11. Modal proportion of the fluids with Raman band positions in MFI in ultramafic granulite (COC17-079a) applied for estimation the residual pressure in the inclusion. Table S12. Modal proportions, mineral composition and phase speciation of stable phases modelled using molecular modelling of MFI_FIB1 bulk composition at both low and high P-T conditions, as indicated with green and yellow stars on Figure 10, respectively. Table S13. Fluid speciation using electrolytic modelling of MFI_FIB1 bulk composition at both low and high P-T conditions, as indicated with green and yellow stars on Figure 10, respectively. For modal proprotions, mineral composition and phase speciation of stable phases see Table 3. Table S14. The results of electrolytic modelling using eclogite bulk composition (sample 97 from Gil  with CO 2 -free and high CO 2 concentrations: modal proportions and mineral compositions of the phases modelled at high P-T conditions (T = 800 C and P = 18 kbar) -a reasonable approximation for entrapment. For details see text Appendix S3. and Figure S1. Composition of the phases are in atoms per formula unit (apfu), except for the fluid phase, whereas wt.% is additionally indicated in columns with F*. Table S15. Fluid speciation using electrolytic modelling of eclogite bulk composition (sample 97 from Gil  with CO 2 -free and high CO 2 concentrations: modal proportions and mineral compositions of the phases modelled at high P-T conditions (T = 800 C and P = 18 kbar) -a reasonable approximation for entrapment. For details see text Appendix S3. and Figure S1.